Exciting Times

Axis Bank’s Axis Aha! Conversational AI powered by Active.Ai wins the DX Leader Award at 2018 IDC Digital Transformation Awards

Categories:

Source: Business Insider

August 16, 2018 – Axis Bank’s AI-led conversational banking assistant – Axis Aha!, powered by Active.Ai, has won in the DX Leader Award category at the 2018 IDC Digital Transformation Awards. Axis Aha! seen a record 3 million plus utterances logged, 97% data accuracy and 40 times growth rate, month on month. The number of new users has increased 31 times while transactions have grown 50x till date.

Triniti, the Conversational AI Engine, powers Axis Aha! with NLP, NLU and NLG which enables it to handle precise context-driven conversations. Currently on web-based platform, it will be on mobile, with popular channels like Facebook Messenger, WhatsApp, and voice-based IoT, Alexa. Convenience of making transactions such as fund transfer, bill payments, top ups and recharges, managing credit cards and more on Axis Aha! is helping the Bank deliver a great Conversational Banking experience to its customers. This disruptive service has contributed to making Axis Bank a leading bank in innovation.

Axis Aha! is selected to be the deserved recipient of DX Leader Award, an accolade reserved for new market entrants or incumbents that demonstrate ecosystem awareness for constant business innovation in the scope of Big Data/Analytics, Cloud, Mobility, IoT, AR/VR to transform products/services, industries, or value propositions. It is awarded by IDC Digital Transformation Awards, which champions organizations that have successfully used digital and disruptive technologies within Asia Pacific.

Praveen Bhatt, Head – Digital Banking & Customer Experience, Axis Bank, said, “Axis Aha! epitomises the Bank’s customer-centric approach. It sets a robust foundation for artificial intelligence based voice and chat enabled ‘Conversational Banking’. Active.Ai is a major contributor to Axis Aha! being awarded DX Leader Award”.

Sharing his thoughts, Shankar Narayanan, COO and Co-Founder of Active.Ai added, “We are honoured to be part of the Axis Aha! journey and proud to see Axis Bank winning the DX Leader Award through our combined hardwork. We hope to continue empower our valued partner, Axis Bank, with the best conversational AI experience!”

Recently, Gartner has named Active.Ai as the Cool Vendor in AI for Fintech in Asia/Pacific 2018. Apart from building the next generation of Conversational Ai experiences for BFSIs, learn how Active.Ai revolutionise their digital strategy and help win awards at www.active.ai.

 

About Active.Ai

Active.AI, a Gartner Cool Vendor in AI for Fintech in Asia/Pacific 2018, is the leading Conversational AI Platform for BSFIs to deliver conversational banking services that help banks redefine their digital strategy. With NLP, NLU and Machine Intelligence, the platform enables natural dialogues through messaging, voice and IOT devices. Founded in 2016, Active.Ai expands its global presence with offices in Singapore, India, Australia and North America. Led by seasoned Fintech entrepreneurs, Ravishankar, Shankar Narayanan and Parikshit Paspulati, with experienced team in AI, Machine Learning, Data Science, Finance, Payments, Banking, Mobile and Design.

Please visit www.active.ai or on twitter@activeaibot

Cognitive collaboration: The emerging role of AI in banking

Categories:

Source: CUInsights

What makes a consumers’ experience exceptional in financial services? The consumer wants every transaction to offer simplified authentication, absolute convenience, and personalized relevance. They also expect superfast transactions and for your financial institution (FI) to leverage their transaction insight to do even more.

To deliver on the promise of an exceptional experience the FI has to address three broad areas:

1. Directly collect relevant transactional data from multiple sources

2. Normalize, analyze, and decipher this data in real-time

3. Convert the deciphered data into information that is actionable

This seems easier said than done because of the sheer magnitude of data being collected. And at times even simple transactions can turn into arduous tasks.

Let me highlight this with an example. I received a letter from my airline telling me that my personal information may have been compromised as a result of a data breach. They assured me that they would do their best to assist me and as a token of goodwill they even offered me two years of “identity monitoring and protection.” All I had to do was to fill out a form online.

As I filled out my personal information into the online form I grew wary of the authenticity of the letter from the airline so I called the airline to make sure the letter was legitimate. It took the representative almost 20 minutes to validate things – from finding the actual letter and then reading the letter to explain the content back to me.

At around the same time, I received a similar notification from my credit card provider. I went online and used voice and text to communicate with the brand. It took me less than three minutes to get the same level of confirmation that took 20 minutes with the airline. And, towards the end of my chat session, I even received a call from a representative to make sure that I was satisfied with the answers.

Many voice and text based chatbots are driven by rules-based engines that are limited in their ability to answer conversational questions. Two of my financial institutions offer me voice and text-based banking services (very rudimentary) but I need to know how to phrase/ask the questions.

The above examples are only the beginning. Emerging solutions are powered by an Artificial Intelligence stack (AI programs) that can deliver seamless and useful experiences. AI can power websites, systems, and especially consumer owned devices (mobile phones, Alexa, Google Home, etc.) to do so much more.

Think about these every day and emerging scenarios:

  • Your vacation takes your family to a new location far away from home. As unfamiliar as you are with the surroundings you commute like a local. The perfect route prescribed through filters – road conditions, weather, and your personal preferences.
  • You summon a vehicle that comes to pick you up and drives away after dropping you off to your destination – you don’t even have to say thank you. There was no driver!
  • A robot delivers pizza to your home, the pizza was produced by another robot, while the vehicle with a full kitchen was driven without a driver. Your dinner, all the way from order to fulfillment was handled my machines.
  • You can visit locations through immersive virtual experiences. Realtors are now using this technology to showcase homes without the buyer even leaving their home.
  • Medical providers are conducting “intelligent” pilots where a medical device monitors a patient’s health and provides real-time suggestions to both physicians and the patient.
  • Food distributors can manage their supply chain by viewing satellite data, predicting weather patterns, and tracking crop yields without even visiting their supplying farms.  
  • 90% of the staff at a hotel chain are robots. They interact with guests, perform most tasks, and ensure measurable performance. The robots speak multiple languages and are extremely perceptive to the guests that they serve.
  • The consumer uses voice commands to direct digitally connected appliances for a variety of tasks. From playing music, to understanding weather, to even directing financial transactions – they use their smart phone, or other digital devices.
  • Artificial intelligence can be strengthened by big data and big machines – today, supercomputers are making available transaction intelligence to brands at a fraction of the cost in real time and can even predict likely outcomes.

Artificial Intelligence is slowly becoming the top tier of the technological ecosystem as we know it. A once unforeseeable technology, AI has made its way through the grassy depths, now emerging as a top predator among other existing technologies. Operating as an intelligent learning system, AI has the ability to learn and grow- using data as an input, to create new and smarter outputs. The financial institution must use AI to make the conversion from blue-collar automation to white-collar automation.

Whether it is driverless cars or a robotic concierge, Artificial Intelligence demonstrates its chameleon ability to implement innovative machinery and also, machinery with a touch of personalization. Financial institutions must face the impending threat of AI on the services they offer, in place of its top predator, payment institutions. Capitalizing on the “sixth sense” of AI, empathy, financial institutions can focus on seven key areas of focus where Artificial Intelligence can be applied:

1. Perfecting Service: Recognizing and treating people well
2. Circumventing Fraud: Protecting people – information, transactions, and data
3. Precise Offers: Serving the consumers what they need, when they need it
4. Optimizing Risk: Dynamic models to balance between success and failure
5. Regulatory Compliance: To ensure that you are doing what you are required to
6. Intuitive Research: Having anytime, easy access to information
7. Empathetic Employees: Allowing employees the time to listen, connect, and serve

These topics harbor the idea of secure, superfast, and personalized transactions, three characteristics that compete with a mounting AI presence. The financial institution is capable of coupling precision with personalization in order to stay in the game.

In a world where artificial is more real and mainstream, new and more intelligent competition has emerged. This competition has the ability to constantly evolve, creating consistently better results, and keeping it ahead of the curb. We should capitalize on these intelligent machines by seamlessly integrating artificiality into our own services, changing the connotation of AI to an environment of cognitive collaboration that is defining the future of financial services.

Personalized Connected Experiences

Read on CUInsight

Forbes India – 6 July 2018

Categories:

Artificial Intelligence + Most Innovative Companies

Since 2007, when Apple launched its first iPhone, everything we knew changed drastically, especially our behaviour in how we consume the world. With just a touch of a finger, the world, or rather, your world is accessible from your palm.

Fast-forward to over ten years after this significant evolution in the world that changed everything, even the financial services migrated from brick and mortar to mobile. The founders of Active.Ai learnt that, the word ‘unreal’ belongs to the vocabulary of yesteryears, and chose to muse the jargons of tomorrow. With significant advances in cloud computing, Artificial Intelligence, telecommunications and smart devices, it is rather a clue for us that industries that touch human lives are getting disrupted by breathtaking technological advances.

Recognising that internet majors, equipped with higher computing capabilities, financial resources, AI leadership and most importantly, engaged customers, Active.Ai believe that the virtualisation of Financial Services will happen sooner than anticipated. This naturally means that incumbent banks and insurance companies need to think and unthink, learn and unlearn, about their brick and advisory networks.

 

Source: Forbes India

View Article

Transforming the Business Banking Experience with AI

Categories:

Source: Medium

As many of the industry analysis reports from McKinsey and BCG over the last few years point out, corporate banks have been the last to join the digital bandwagon. Corporate internet and mobile banking took off to some extent over the last decade. Straight through processing (STP) for various transactions through direct integrations between the ERPs and the banking systems has eased out a lot of friction in the business banking processes. However, the overall digitalisation is way less than the other lines of businesses in banking.

Even though many areas of corporate banking have undergone some amount of digital transformation, when it comes to customer experience, there is still a lot of dependence on the RM, even for simple queries starting from balance checks to financial transaction status. Such engagements serve the immediate requirement of the client but are of no real “value added” from a costly channel like the RM.

These engagements extend to contact centers through calls and emails which are all extremely high cost and high turn-around-time channels. The interactions are not exactly delightful when the client has to either wait in a call queue or be kept on hold for the agent to get the appropriate details to answer the query.

Most RMs engage with their clients through calls and other messaging channels which are outside the institutional purview. Thus, these interactions are not recorded in any way. Adding to that is the security consideration of such exchanges. The only form of authentication is the RM’s belief that he or she is speaking to the actual client on the other end of the channel. It is also well known that the attrition in the RM community is very high and along with the RM, all the insights on the clients, which should have been added to the institutional knowledge, also get lost to the institution.

Now cut to retail banking which has been at the forefront of digitalization. While “mobile first” was the mantra a few years back, today it is “AI first”. As various reports point out, AI investments in banks are the highest in the area of customer experience. We are seeing a big movement in retail banking customer experience where the interaction is moving from structured, form-based channels like internet banking and mobile applications, to unstructured and natural-language-based conversational engagements. There are multiple channels that the banks are opening up, ranging from chatbots on their websites and mobile applications, to going where the customers are, like Facebook messenger, WhatsApp and so on, to voice channels like Alexa and Google Assistant.

As mentioned earlier, corporate banks were always more into unstructured, natural language interactions. However, through these high cost channels and at times, channels that the institution has no insight into and with human authentication only, it ends up with no records. Moving to another conversational channel, which can be even lower cost than mobile and internet banking and available 24/7, can be the next big digital transformation for banks to serve their enterprise clients.

With AI based conversational banking, the banks can reach out to the business clients in a channel which they are already using. Most of them use the channel for their internal collaboration or to interact with their RMs, which means that the institutional data and knowledge will be saved. This will be in addition to internet banking and mobile banking application today. However, over a period of time, the conversational channels can see a much higher adoption. The AI based conversational channel should have the ability to seamlessly handover to the RM or a human agent when it cannot cater to the customer request, to ensure the service levels remain at a high level even in the early days when the system has had little training. In fact, the best experiences can be provided by such human-machine combinations instead of any one of them.

Conversational channels can be deployed as extensions of the website, internet banking and mobile banking, through additional channels like Skype, Lync, to name a few, that businesses use for their internal collaborations. Banking through Skype, Lync type channels by business clients essentially mean that the corporate users don’t have to open a different website or mobile application but instead continue with their banking from where they are collaborating with their colleagues for other purposes.

Besides these messenger channels, for immediate, short and critical queries of the CXOs, banks can create conversational experiences through voice channels like Alexa, GA and so on.

Conversational channels in business banking are relatively unexplored areas. On that note, it has great potential only if the channel specific customer experience (CX) design is completely reimagined for the channel and not replicated from web or mobile. Even the voice interface design (VUIs) should be differently thought of from the messenger conversations. Only then will the adoptions of these channels soar.

The conversation experience design should be a combination of the target persona in the corporate and the channel and must start from the requirements of the persona. A Juniper report shows that there is an average of four minutes savings per call when transactions are done over chatbot. In addition to that, a Medici report shows that 70% of customers today prefer a messenger channel over a call. Our belief is that a well-designed conversation engagement through any channel should bring down the time further and provide a delightful experience. The entire experience design should be on an extremely strong backbone of conversational AI, whose (1) natural language understanding (NLU) should be able to understand the nuances of complex human conversation and the corporate banking domain ontology, and (2) Machine Learning capabilities keep making the conversations more individualized to the corporate and the particular user.

One of the possible ways to start off the conversational channels for business banking is to analyze the high-volume interactions in the contact centers and with the RMs. While the contact center data will be easier to assimilate, the RM data points are equally important since that is probably the area that needs to be brought into the institutional purview asap. Such analysis will not only help the banks prioritize the type of interactions to provide in this channel, but the actual conversation data can be used for training the AI system on the utterances of the customers and the possible replies the bot can provide.

Providing a conversational channel for prospects (who are not yet customers of the bank) could be the very first step for business banking. This could be a completely new channel on the bank’s website to help out the prospects with products that they are seeking or providing the simplest and fastest way to frequently asked questions (FAQ). The experience can be truly delightful only if the AI engine understands the queries exactly and the replies can be given to the point with minimal redirection to elaborate documents. Wherever the queries are not clearly understood, a seamless and quick handover to a live agent could help keep the experience frictionless. This combination of AI based conversational channel and live agent, only in extreme cases, could be a great recipe for client conversion instead of an abandonment.

FAQs could be a good start for the conversational channel for existing customers as well. A lot of their generic queries do land up with the RMs and call centers. Having a simple self-help in their corporate messenger could be an immense help. The next phase could be to take up service requests, such as statements and certificates, and transaction approvals, like corporate multi-user approvals for high value transactions, through this channel. High volume financial inquiries like remittance status, beneficiary credit, invoice payment stage and so on could be the subsequent phase of the conversation channel. Based on the adoption, the conversation channel could provide all the high-volume interactions over a period of time.

Banks can also look at virtual assistants as a part of business internet banking that could be completely conversational. Corporate internet banking with its multitude of options could overwhelm the user and drive her to the call center. The conversational assistants can be invoked within the internet banking session that can help her navigate through the system or help her with a balance inquiry when she is in the middle of approving an ad-hoc payment without having to move out of the approval option. This can prevent that costly call to the bank and save her a lot of time as well.

In summary, we see that conversational channels based on powerful natural language capabilities and continuous machine learning can help banks provide their business customers a delightful, always available experience and reduce the usage of costly and cumbersome channels. That will essentially mean that they reach out to RMs and call centers for real value adds like advice on new facilities or better deployment of funds that can be more revenue generating for the bank… and a fringe benefit could be higher retention of RMs.

To know more on how to transform banking experience, please visit Active.Ai.

Source of References: BCGMcKinseyGoMedici

Read on Medium

Staying Sane and Optimistic amid the AI Ballyhoo!

Categories:

Source: Medium

At NIPS 2016, there was an unprecedented story building up. Something that got every AI enthusiast agog about an unknown AI startup ‘Rocket AI’.

The names associated with the hot startup were pioneers in the AI field and it was informed to the media that there was major announcement soon to come. There was even a workshop held, where one of the researchers explained about the concept of Temporal Recurrent Optimal Learning to the house full of researchers and media personnel.

The whole community was abuzz with the jargons “Temporally Recurrent Optimal Learning”, “Jacobian Optimized Kernel Expansion”, “Fully Automatic Kernel Expansion”, a couple of which were coined by a leading AI researcher. These made rounds on web with a hype so strong that it got 5 Major VCs reaching out to them for investment.

There were rumours about Rocket AI’s acquisition as well — all within a day.

Only to be figured out, later, to be a joke!

Temporally Recurrent Optimal Learning = TROL
Jacobian Optimized Kernel Expansion = JOKE
Fully Automatic Kernel Expansion = FAKE

If you still don’t get the joke, you should probably get yourself the highly coveted RocketAI T-shirt @$22.99 a piece.

 

The Rocket AI launch story at NIPS16 was a great lesson on why it is imperative to disentangle noise/hype from the real advancements around a technology on surge.

But let’s be honest, can we inculpate anyone when they fall for the AI ballyhoo? Such hype has always been associated to technologies with potential to change the world. In many ways, it already is. Amidst the extravagant hype, there has been multitudinous success stories as well. Autonomous vehicles, AlphaGo victory over world №1 Go player, OpenAI DOTA2 bot challenging top professional gamers, surge in successful applications in healthcare & fashion owing to advancements in computer vision, is a testimony to the fact that AI has arrived and it’s here to stay.

Of late, there has been a myriad of NLU/NLI/QA datasets crowdsourced and released for the research community. It has undoubtedly propelled the DL research efforts for NLP applications. BABI inspired multiple forms of memory networks, SQUAD/newsQA impelled BiDAF, AOA, mnemonic reader, fusion-net etc, SNLI/MultiNLI,RTE exhorted multitude of attention networks.

However, an agonising trend that concerns me is the over-reliance of young AI practitioners on SOTA (State-of-the-Art) for identical problems. Whether you’re a researcher involved in primary research on AI, a product developer utilising the existing frameworks/algorithms to a given business problem, or an executive evaluating a product, one should always be mindful of the capabilities and limitations of any such SOTA, and its applicability to the given use-case/dataset.

After all, “There are no free lunches in AI”.

Lets have a look at a few observations around the issues with a few popular Datasets used for benchmarking NLU/NLI algorithms and the State of art solutions for them.

 

Issues with the popular State-of-the-Art solutions and the datasets

1. Squad performance State-of-the-Art models dropped drastically after adversarial examples were added in V2.0 compared to V1.1.

2. Best performing model on Squad dataset only yields close to 40% f1 Score on NewsQA testSet.

3. InferSent seems to rely on word level heuristics for high performance, for SNLI challenge. Most of the contradictory sentences have no overlap in words and with high overlaps, it is more likely to be entailed. This indicates InferSent may be simply learning the heuristic that high overlap could mean ‘entailment’, presence of negation could mean ‘contradiction’. For pairs with high overlap between words with the only token difference among the pair being an antonym, they are classified as ‘contradictions’. Presence of an additional non-antonym word throws the predictions off, classifying it as ‘entailment’.

4. Try training a fastText classifier on SNLI dataset and ensure that the model doesn’t get to see the premise but only observes and predicts based on the hypothesis. Be ready to be baffled, because the prediction accuracy would be way higher than some of the baselines! A task which makes no sense to be solved without considering both hypothesis and premise together, it’s quite amusing to find a model doing well without the premise.

5. Quantifying determiners and superlatives were mostly included to make the sentences look similar but always present a deviation from the hypothesis. Thus, a ‘contradiction’ in most of the cases.

6. Negation, presence of hypernym and hyponym, and overlap of words largely contributed to the entailment class.

These unprecedented benchmarks (SNLI inference problem: train accuracy 95%, test accuracy 91%) might give an illusion that natural language inference is already a solved problem. However, these evidences say otherwise.

Tesla AI director Andrej Karpathy showed with few simple examples of how a deep-learning model could be fooled, by adding little amount of noise.

Panda gets recognised as gibbon with high confidence.

The Graph above shows the computation requirements of different AI solutions against their timeline of existence. Is it just the compute power & more data or better algorithms?

Coursera co-founder & Google brain founding member Andrew NG claimed the AI systems could diagnose diseases from X-ray reports way better than radiologists.

The results later proved AI still has a long way to go, in making radiologists totally obsolete.

 

 

Some interesting excerpts from The former Facebook AI head Yann Lecunn’s interview with Spectrum.

 

Research frontiers

Lets look at some of the recent AI breakthroughs in computer vision and NLU. Nvidia, OpenAI, Microsoft and Google have been the forerunners.

Below are the image outputs from the OpenAI model which performs Mix-match and Image manipulation. The output is extra-ordinarily high definition and seems natural, although they aren’t real human beings. The third image shows how Nvidia’s model re-creates an unrecognisable object in an extremely blur image automatically.

 

Source: https://blog.openai.com/glow/ (Image manipulation and Mix-match)

 

Source: https://www.dpreview.com/news/0229957644/nvidia-researchers-ai-grain-noise-images-photos

 

Learning with lesser data: Meta learning Meta learning is the process of learning to learn. A meta learning considers a distribution of tasks, and generalises to learn a task efficiently with a very less amount of data. Reptile, MAML are a few popular Meta learning algorithms. Most of the Machine learning usecases at present, are dependent upon supervised approach of Learning, requires a lot of annotated data which is scarce to find. Breakthrough in meta learning approaches can really pave the way to building learning systems which can learn with the very less amount of data, just like how humans learn with substantially small amount of observations.

Automated annotation: There has been a lot of progress in this front, with the evolution of powerful sampling strategies and prediction reliability estimates which can help in sampling a minimum set of instances required to be annotated for the model to approximate a function good enough to replicate the behaviour of the model trained on the full dataset. These frameworks can be applied to automate the annotation process by reducing the manual efforts to a great extent.

Multitask learning: Can you cast each of these NLP tasks as question-answering task and solve it with one network? Question answering, machine translation, summarisation, natural language inference, sentiment analysis, semantic role labelling, zero-shot relation extraction, goal-oriented dialogue, semantic parsing, and common sense pronoun resolution?

Evaluation framework to identify annotation biases: The issues of annotation bias is very much evident by the issues identified in some of the well known public corpuses used for research purpose (SQUAD 1.1, SNLI ). A lot of the existing State-of-the-Art algorithms may not perform high on these datasets if these annotation biases are removed, which will only entice the research community to push the limits and explore better approaches to tackle these problems.

Unsupervised approaches of learning: There has been a lot of research in supervised learning approaches, and its not an unknown fact that given a sizeable training data, it’s no more difficult to train a system to do a varied activities of prediction/recognition/generation etc. The data gathering, annotation, curation is an expansive process and for a lot of real-world problems this becomes a bottleneck.

Unsupervised Sentiment Neuron, Unsupervised Language Modelling using transformers and Unsupervised Pre-training, yielded encouraging results.

Contextual and efficient word representations. Word2vec , Glove & Cove were a good start. FastText made it better, ELmo seems to have surpassed the others based on its performance over a lot of NLU/NLI tasks. The contextual word representation for the words have proved to be quite helpful in downstream tasks.

We are starting to see impressive results in natural language processing with Deep Learning augmented with a memory module. These systems are based on the idea of representing words and sentences with continuous vectors, transforming these vectors through layers of a deep architecture, and storing them in a kind of associative memory. This works very well for question-answering and for language translation

Search for Better optimisers: Using machine learning to find better optimisers proved to be quite beneficial. Optimisers play a pivotal role in the performance of Deep learning architectures. Some of the most commonly used optimisers being Adam, AdaGrad, SGD etc. The search for efficient optimisers is just as important as discovering better optimisation algorithms. Google AI showed that reinforcement learning can be utilised for searching better optimisers for deep learning architectures. PowerSign and AdaSign were discovered using the same, and have proved to quite efficient for many DL usecases.

Zero-shot learning/One-shot learning: Can a supervised learning model be trained to predict a class that is not present or are entirely removed from the training data?

 

Interpretation vs Accuracy

An interesting debate surfaced up at NIPS 2017, with Ali Rahimi and Yann LeCun locking horns over Rahimi’s remarks of Machine learning becoming the ‘alchemy’. Blaming the present Deep learning systems to be a black box or at best an experimental science, Ali maintained the AI systems need to be based on verifiable, rigorous, thorough knowledge, and not on alchemy. LeCun argued that the “lack of clear explanations does not affect the ability of deep learning to solve problems!” While LRP(layer wise relevance propagation), deep Taylor series and LIME (Local Interpretable Model-Agnostic Explanations) are a few methods being utilised to make these systems explain their predictions, it is still in its nascent stage. Or do we need to settle this debate by saying, “It might just be part of the nature of intelligence that only some part of it is exposed to rational explanation. Some of it is just instinctual or subconscious, or inscrutable.”

The voice in favour of supposedly imminent AI winter is just as strong as the ardent optimists who believe AGI is just around the corner. Although the later may not be likely anytime soon, The former is more likely if AI breakthroughs are overhyped beyond truth.

“AI researchers, down in the trenches, have to strike a delicate balance: Be optimistic about what you can achieve, but don’t oversell what you can do. Point out how difficult your job is, but don’t make it sound hopeless. You need to be honest with your funders, sponsors, and employers, with your peers and colleagues, with the public, and with yourself. It is difficult when there is a lot of uncertainty about future progress, and when less honest or more self-deluded people make wild claims of future success. That’s why we don’t like hype: It is made by people who are either dishonest or self-deluded, and makes the life of serious and honest scientists considerably more difficult.” Yann Lecunn

 

Ashish is the VP of AI Product for Active.Ai. Find out what the future holds for Conversational AI at www.active.ai

Read on Medium

Conversation Design and breaking from shackles of App mindset

Categories:

Source: Medium

Conversational Interfaces are relatively new. Facebook was one of the earliest to popularise it and this was a little over 2 years ago. The last 4 decades have seen human computer interfaces stabilise around point-and-click GUIs evolving all the way from Apple Macintosh introduced in 1982 to the iPhone (that introduced multitouch and swipes to the masses) launch in 2007. Massive leaps has been made in those 4 decades in defining usability standards and guidelines for all platforms like desktop, web and mobile applications. Conversational interfaces require a whole new mindset to realise its potential to the fullest.

As you read the rest of this article, you will find that a lot of examples are provided in banking context, which is where most of our experience and audience lies. But I see no reason why most of these principles cannot be applied to pizza ordering, for example.

Application interfaces had one requirement that drives its design and all usability guidelines around it. If you make an application module, for example, to do a bill payment, the best way to manage this is to automatically present the options on screen. Even those requirement settings that rarely changed, like for example, the source account for making payment, or the payment date set as default to a certain day and time, or even optional elements, like comment.

What defines the application design is the fact that, unless the options were presented on screen, it cannot be used. Users have also developed a sense of selective blindness overtime as they got more familiar with the interfaces. So they come online, just select the biller, enter the amount, just glance if the other default fields are correct and complete the transaction.

Moving to conversations. The first instinct of an experience designer is to look at current design of mobile/internet banking apps and design the conversation based on that. That is one of the most terrible way to start. Once you have seen the mobile/internet banking screens, it is very difficult to un-see it. We have some cases where a complete form in an internet banking experience is replicated into the conversation.

This is what we say to all our customers. When designing a conversation for any given use case, close your eyes, imagine your smart and intelligent secretary/spouse just walked into the room and you asked him/her to remit the rent. He/she already knows who your landlord is and how much the rent is. The response you would expect is “Sure, I will do that right away. Can I get you a coffee or anything else?”

Now imagine he/she asks “To whom do you want to pay? How much? From which account? When?” You will probably look for another secretary. If he/she is your spouse, too bad. 🙂

The point is, your bank, unlike most organisations on this planet (save for Google and Facebook), knows you the most. Replicating that secretary scenario on a conversational channel is not even AI, it just requires simple historical look up against users’ past transactions. Yet, so few are thinking along such lines.

The guideline for a successful conversation is this. If the user cannot do it faster on a conversational channel compare to a mobile application, they wont don’t do it again. I transfer funds regularly to my wife for our expenses and measured the time it take to complete the activity on the conversation platform. It is definitely faster.

1. Go to bank.com
2. Open the bot
3. Type “Transfer 50000 to wife”
4. Authenticate
5. Confirm
6. Enter One-Time password (OTP) received via SMS
7. Done.

This is nearly perfect for a conversational channel. One small change I would make is to avoid the confirmation step if this is a regular pattern of the customer and jump directly to “Ok. I am transferring 5000 to your wife. Please enter the OTP to proceed.” If this is on another registered channel like Facebook Messenger, the authentication step can be skipped as well.

The idea is to design conversation for pure audio channel from the start. That keeps you grounded. Facebook did a great job in defining standard templates. The purpose of those templates such as cards, lists and carousels is to provide information and avoid typing as much as possible. Opening a design tool like Sketch/Balsamiq to design the conversation flow for a usecase is a sure shot way to derail the thought process and may end up too app-like.

Conversation designer should be more like a playwright than a UX designer. Once the playwriting process is complete, you can then add lists and cards at appropriate places within the flow if the channels allow it.

Another key aspect of designing good conversational applications is the 80–20 rule and ability for users to change the assumptions. I will illustrate via an example below from a recent experience.

Recently when we designed a flow for opening a term deposit with a bank, one thing lead to another and the proposed flow ended up with 15 steps that was replicated based on its other channels.

It goes like this:

1. How much you want to deposit
2. One time / Recurring
3. What Tenure
4. What happens on maturity… and so on.

One of our co-founders threw a challenge. Why can’t we do it in one step?

And this is how we did it:

1. What is the information that we really need from the user? It is just the amount that he wants to deposit.

2.Now the bank knows that 80% of the users who do recurring deposit, the amount is less than INR20,000. Therefore, it is quite fair to assume recurring or single deposit based on that amount.

3. Now, the bank also knows that for a recurring deposit, what the most common tenure chosen by 80% of the users will be or if it is one time, the bank can assume the tenure is to get the best interest rate possible.

4. The bank also knows what the behaviour will be for the majority upon maturity.

So, essentially, the flow goes like this.

User: Open a deposit for 50k

Bank: Sure. The best I can offer you is 7.6% for a 11 month term. Proceeds will be credited back into your account after 11 months. Interest of INR 767 be credited quarterly into your account . Shall I proceed?

User: Yes, Please.

In Summary:

Design the conversation so that it is extremely fast for 80% of the users. Its not fair dragging the conversation just to cater for the other 20% outliers.

Now comes the beauty of conversational channel. Although the flow was designed with the 80% in mind, it does not mean we are ignoring the remaining 20%.

As an example, those 20% of users can simply ask “Can you make the tenure 5 months please?” and customise any assumption that the system made for them. And over time, if a user always tends to change his tenure to 5 months, the conversation system should start assuming 5 months tenure for that particular user. Or if the user has no fixed pattern of deciding tenure, it makes sense for the bot to ask for the tenure every time.

This is true personalisation. That is what your secretary/spouse would do. This allows the conversation system to be extremely fast for the 80% on day 1 going live. And it evolves automatically for the remaining 20% to be fast for them too, based on their patterns.

Before signing off, I would like to highlight that not every usecase fits well into a conversational interface. Apps and websites will not fade away. Try not to fit every usecase into a conversational journey. Any usecase that is too long (like opening an account and begin a brand new relationship with a bank, or apply for a home mortgage) are not naturally suited for conversations. Although conversations is the in thing now, users are better off filling a form on a website or mobile app where he can go back and forth, changing inputs etc.

Conversational interfaces provide a very powerful way to interact with your users. Design the system that can leverage the knowledge you have about your customer base and create journeys that wow them.

Read on Medium

Triniti meets Oracle

Categories:

I’ve been working with Active.Ai for more than a year now, leading operations and product for the North American market. I focus on identifying product requirements unique to the target market, working with our clients, prospects, analysts and gathering market research information.

Active.Ai’s key difference is our team – I couldn’t ask for a more passionate, talented, helpful and smart bunch of colleagues. Together, we have built a platform, Triniti, that’s at least a year or two ahead of the market place in terms of allowing a natural conversation with an AI system. Our focus on Financial Institutions and their needs has paid off as we have produced pre-built user journeys and datasets that accelerate project development and deployment.

When Oracle Banking team invited us to join them for the Industry Connect event in New York City, we jumped at the opportunity as we share the same ethos towards creating a secure and scalable banking platform. We believe that we complement Oracle’s core banking systems with the front end NLP capabilities for next-generation user interfaces.

For a demo day like this, we prepare by profiling the audience – is it more technical or more business oriented? Based on this, we draw up a presentation deck that addresses the relevant features and benefits of our platform. We include existing collaterals, and if time allows, we include custom demo materials. We made sure that every bit of Triniti’s capabilities in delivering top conversational experience is showcased.

At Industry Connect, Oracle provided us with an opportunity to work together with them in a ‘hackathon’ format. This was hectic but fun, and the end result is there to see in the video – a conversational website utilizing Oracle Banking APIs.

Madhav Mehra is the VP Product & Operations for Active.Ai for the North American market. Find out more on how you can ride along their journey into the future of Conversational AI at www.active.ai

Citi GPS Bank Future – March 2018

Categories:

 

Q: Tell us about the genesis of Active.Ai?

We started the company in early 2016. I did a trip to China in mid-2015, looking at how financial services were being delivered, and knew China was making huge strides in using AI and the latest technologies to enable financial services. It was fascinating to see how China got there. By the end of 2015, it was clear that the “mobile first” model for financial services, built from China, was going to be disruptive and take on the world.

View Article

How do we anticipate the future?

Categories:

The last few years have seen significant advances in cloud computing, AI, telecommunications and smart devices. Every industry which touches human lives is getting disrupted for the better by technological advances at a breathtaking pace.

At Active.Ai, our team has been imagining the future of financial services at our skunk works style innovation hubs.

This helps us bring ideas from experimentation to proof of concepts to pilots at great pace.

I recently had a peep into our own Matrix for wealth management, where I walked around a mocked up virtual portfolio, pinched and moved graphs across asset classes, wearing a Holo Lens. It felt like so much like science fiction, that I had to pinch myself!

We can very well imagine a future, where an Advisor and Client are engaged over conversations, on such platforms. Banks can create virtual branches and tellers. Insurance companies can be engaging in virtual claims discussions. Robo-advisors can add personas.

Yes, it seems unreal but we can see that in near future, these capabilities may become mainstream in customer engagement. 2007 was the launch of the smartphone (Apple’s iPhone). In a decade, everything changed and financial services moved from Brick and Mortar to mobile.

Given that Internet majors, having tremendous computing capabilities, financial resources, AI leadership and most importantly engaged customers, we believe that the Financial Services virtualization will happen sooner than anticipated. Incumbent banks and Insurance companies will need to think (or rather, ‘unthink’) about their brick and advisor networks.

Maybe in the year 2021 you would be greeting your banker wearing smartglasses in a virtual branch.

Do you see this possible in the future? Tell us what you think: hello@active.ai

 

Source: Medium

How Chatbots in Financial Services are Evolving with NLP and NLU

Categories:

AI-powered chatbots can understand your specific banking needs using NLU — and also detect how you feel and predict what you’ll say next.

Processing human language is a challenge for artificial intelligence — the way we converse is nonlinear, irregular, emotional, and full of context. The goal is for AI to hold a back-and-forth conversation with a human in a way that feels natural — despite the fact that it’s with a machine. Natural language processing — or NLP — is the first step to making this a reality.

In customer service-focused industries, like banking, chatbots equipped with NLP can analyze, process, and communicate with users, using language they understand. NLP techniques categorize customer data by tagging parts of speech, correcting spelling, and re-formatting numbers and dates into something the machine can read.Read more…

The Best 50 Cities for a Startup in the World

Categories:

Source: Valuer.Ai

Singapore has a lot to offer, from modern infrastructure and an educated workforce to a strategic proximity to many emerging and developed markets. Outranking Silicon Valley as the indisputable top and startup capital, which may come as a surprise.

Let’s look back into its history.

Remember the recent competition between Singapore and Hong Kong? Singapore won by earning recognition among investors as the best place for expanding or starting a new business in Asia.

The two are former British colonies, with well-structured governments, free-port trade to foreign investors.
Both depend on economic growth.
The two have a city megapolis and reach between 5-8 million citizens.
Both expanded after World War II, surpassing even Japan’s GDP per capita.
Today Singapore has one of the best education systems in the world and has become a startup hub for entrepreneurs, tech companies and investors.

Here is a list of promising startups in the area: CoinPip, Datarama, Greyloft, Honestbee, Mighty Bear Games, Nugit, oBike, ShopBack, Spark Systems, Chope, 99.co, Active.Ai

View Article

Digital Davids vs Financial Goliaths

Categories:

Active.Ai Ravi Digital Davids vs Financial Goliaths
We live in an interesting time of disruptions. Like many industries, the financial services sector is facing its Kodak moment. The incumbents have the experience, the challengers are creating the experience. Here are a few of the many moments that I feel will define the next decade for the incumbents.

Read more

Finacleconnect – March 2018

Categories:

Shankar Narayanan: I think blockchain will definitely be one of the key technologies that have a significant
impact on banking this year. In addition to areas like trade finance and remittances where there is a lot of traction already, I think blockchain will also make a difference in the know-your-customer process bringing in a lot of efficiency and transparency. Secondly, regulations or not, banks will focus on setting up open banking framework to play a larger role in the banking ecosystem. This will be important for banks to develop a new revenue stream and increase their reach through third parties. Thirdly, Artificial Intelligence cannot be ignored any more in banking. 2018 will see more of conversational banking services through chat and voice based interfaces across their digital channels.

View Article

CIO India Review – March 2018 Issue

Categories:

As we gaze far into the year, we can see that the conversational AI landscape is primed for increased consumer adoption. In fact, in a recent survey, nine out of 10 people said they prefer messaging directly with a brand. This year, Apple, Facebook, Google, and Amazon, all leaned into messaging and conversation. While chatbots are still at a nascent stage in the banking industry, bots will quickly gain in sophistication to the point that they will be able to perform all tasks previously owned by customer service representatives. Active.ai, a Singapore based Fintech startup with an innovation lab in Bengaluru, is using artificial intelligence (AI) to deliver Conversational banking services. The startup helps banks redefine their digital strategy for the future, bringing in automation and insightful customer engagement. Built for banking technology, their conversational AI uses advanced natural language processing (NLP), natural language understanding (NLU) and machine intelligence to enable customers to have natural dialogues over messaging, voice or IOT devices.

View Article

Singapore-based Conversational AI Startup Breaks into US Market with Key Bank Clients

Categories:

Active.Ai kicks off a roadmap for US growth with two new bank clients.

February 27, 2018 (New York, NY)— Singapore-based artificial intelligence startup, Active Intelligence Pte Ltd, made a significant step into the United States by signing two important clients.

An enterprise platform for financial services that helps facilitate intelligent micro-conversations, Active.Ai, raised a total funding of US$11.75m from marquee investors, already seen notable success enabling four of the largest banks in India, along with other top-tier banks, and insurance companies in Malaysia, Singapore, Thailand, and the Philippines. Its financial services have span into other verticals like wealth management and securities trading. On the heels of announcing projects with Axis Bank and CIMB Bank being live, Active.Ai now have their sights set on the US market, and already the 24-month-old company has secured a strong foothold in the Americas.

Elizabeth Duke, SVP of Business Development, will be responsible for driving new business opportunities in the US. Duke shared “I’m excited to be part of a growing team that’s focused and committed to the US market. With significant bank customers across Asia, it’s clear the Active.Ai team is ready to deploy at scale in the Americas. The potential is huge to up-end the level of service banks can deliver to their clients”. Duke brings 30 years of global sales and marketing experience in financial services, including senior-level positions at MasterCard, CSAM, Carta Worldwide, and Network for Electronic Transfers Singapore (NETS). She will play a key role in growing the company’s presence across North America alongside co-founder and CEO, Ravishankar, and co-founder and COO, Shankar Narayanan.

Ravishankar, explains: “We’ve already seen a strong interest in our full-stack solution, and these early wins suggest US financial services are ready to incorporate an innovative full-stack AI solution.” He adds, “As Active.Ai scales and enters new markets like the United States, we continue to hire talented people to drive product innovation and growth—2018 will be an exciting year for banks in the Americas.”

Active.Ai runs on a proprietary artificial intelligence engine called Triniti that enables financial institutions to have a meaningful, intuitive engagement with their customers across multiple apertures, using messaging, voice, and IOT devices. This unique solution—built from the ground up by Active.Ai—includes Machine Learning, Natural Language Processing, and Natural Language Generation. It’s arriving just in time for the industry. Shankar Narayanan explains: “The financial services and insurance industries are evolving quickly to remain relevant to changing customer expectations. Using advanced conversational AI, we see an exciting opportunity for these companies to create a natural dialogue and more meaningful connections with their customers using everyday micro-conversations.

Download PDF

Active.Ai raises $8.25 m Series A funding, led by Vertex Ventures

Categories:

Singapore and Bangalore, India, Nov 13, 2017

Active.Ai, a Singapore Headquartered Fintech platform with an innovation lab in Bengaluru, that delivers conversational banking through artificial intelligence (AI), announced a US$ 8.25 million Series A financing led by Vertex Ventures, Creditease Holdings and Dream Incubator. Existing investors Kalaari and IDG Ventures India will also participate in the round. Ben Mathias, Managing Partner at Vertex Ventures and Vani Kola, Managing Director at Kalaari Capital will join Sanat Rao from IDG Ventures India on the Company’s board of directors. Anju Patwardhan from Creditease will join as a board observer.

Active.Ai’s proprietary AI engine, Triniti enables financial institutions to have a meaningful engagement with their customers in an intuitive natural format over multiple apertures covering messaging, voice and IOT devices. This full-stack solution has been built ground-up and comprises Machine Learning, Natural Language Processing and Natural Language Generation. Keeping in mind the requirements of financial institutions, the company offers flexibility of deployment: on-premise or in the cloud.

Founded in early 2016, the company is working with some of the top-tier Banks and Insurance companies in India, Malaysia, Singapore and North America and is planning to expand into other verticals like Wealth Management and Securities Trading. The company’s vision is to become the AI platform of choice for leading Financial Services companies across the world.

Ravi Shankar, co-founder & CEO of Active Intelligence said:  “I am excited by the future possibilities of AI and how this technology will shape the banking and financial world. There is very strong need for banks and financial institutions to evolve fast and empower customers with the ability to do transactions as part of their habitual daily micro-conversations. With the fresh injection of funds, Active.Ai will scale up and continue to hire talented people for the AI team and focus on building the best enterprise product in the market.”

Ben Mathias from Vertex Ventures said: “I am very excited to partner with the Active.Ai team to redefine the paradigm of customer experience in financial services. The threat of fundamental disruption is very real in the financial services space, and companies such as Active.Ai are making it easy for incumbents to not just remain relevant, but to get ahead in an ever evolving market. At Vertex, we have been strong believers in Artificial intelligence helping solve problems specific to industry verticals, and that has been a key driver of our decision to partner with Active.Ai.”

Vani Kola, Managing Director at Kalaari Capital said “We are very excited to have Vertex, Creditease and Dream Incubator join us in this exciting journey of building a world class AI product from India. The fresh round of funding will help us invest even more into technology and expand our footprint into other geographies.”

“Active.Ai’s NLP platform for conversational banking is rapidly gaining acceptance with key banking customers globally.  We continue to be impressed with the passion and capabilities of the team, and are excited to be part of the next phase of their journey.” said Sanat Rao, Partner, IDG Ventures India

Download PDF

Renaissance era for Banking

Categories:

 

“The World needs Banking, but it does not need Banks”

– Bill Gates

Today, we are living in the “Renaissance era of Banking” and this transformation that we see is from disruptive changes across regulations, technology, and the very manner in which banking is consumed as a service. The opportunity to transform banking is now a top agenda for most nations globally as they seek to make banking inclusive and accessible to billions.Read more…

Chat Banking: The Path to Ubiquitous Mobile Banking

Categories:

chat banking GUI

Menus, icons, and clicks from the world of GUI have served us well by bringing computing (and so Online Banking) to the masses. In today’s hyper mobile world, the ‘gooey’ model is crumbling under the weight of the increasing number of services and amount of information people want available on their mobile devices.

Natural User Interface (NUI) is an interesting promise to make GUI unobtrusive and invisible by allowing natural human behaviour (touch, gesture, and speech) to interact with devices. But we don’t yet feel so natural interacting with machines for general use. Google Glass, Oculus Rift, and Leap Motion won’t become widespread anytime soon.

Enter Chat UI. Texting is hot for chatting among friends. Add to it some richness from the world of GUI and it may soon become the primary way in which people bank (and shop, and search even) on their mobile devices.

Read more…

Digital Bank: Why Human Capital strategy matters most

Categories:

Preparing your bank to go digital is not only about rethinking your technology, it’s about rethinking your bank, your customer engagement and most importantly your People Capital.

The lost opportunity for state owned banks: Flash Back: 1985-2000.Read more…

Axis Bank’s Axis Aha! Conversational AI powered by Active.Ai wins the DX Leader Award at 2018 IDC Digital Transformation Awards

Categories:

Source: Business Insider

August 16, 2018 – Axis Bank’s AI-led conversational banking assistant – Axis Aha!, powered by Active.Ai, has won in the DX Leader Award category at the 2018 IDC Digital Transformation Awards. Axis Aha! seen a record 3 million plus utterances logged, 97% data accuracy and 40 times growth rate, month on month. The number of new users has increased 31 times while transactions have grown 50x till date.

Triniti, the Conversational AI Engine, powers Axis Aha! with NLP, NLU and NLG which enables it to handle precise context-driven conversations. Currently on web-based platform, it will be on mobile, with popular channels like Facebook Messenger, WhatsApp, and voice-based IoT, Alexa. Convenience of making transactions such as fund transfer, bill payments, top ups and recharges, managing credit cards and more on Axis Aha! is helping the Bank deliver a great Conversational Banking experience to its customers. This disruptive service has contributed to making Axis Bank a leading bank in innovation.

Axis Aha! is selected to be the deserved recipient of DX Leader Award, an accolade reserved for new market entrants or incumbents that demonstrate ecosystem awareness for constant business innovation in the scope of Big Data/Analytics, Cloud, Mobility, IoT, AR/VR to transform products/services, industries, or value propositions. It is awarded by IDC Digital Transformation Awards, which champions organizations that have successfully used digital and disruptive technologies within Asia Pacific.

Praveen Bhatt, Head – Digital Banking & Customer Experience, Axis Bank, said, “Axis Aha! epitomises the Bank’s customer-centric approach. It sets a robust foundation for artificial intelligence based voice and chat enabled ‘Conversational Banking’. Active.Ai is a major contributor to Axis Aha! being awarded DX Leader Award”.

Sharing his thoughts, Shankar Narayanan, COO and Co-Founder of Active.Ai added, “We are honoured to be part of the Axis Aha! journey and proud to see Axis Bank winning the DX Leader Award through our combined hardwork. We hope to continue empower our valued partner, Axis Bank, with the best conversational AI experience!”

Recently, Gartner has named Active.Ai as the Cool Vendor in AI for Fintech in Asia/Pacific 2018. Apart from building the next generation of Conversational Ai experiences for BFSIs, learn how Active.Ai revolutionise their digital strategy and help win awards at www.active.ai.

 

About Active.Ai

Active.AI, a Gartner Cool Vendor in AI for Fintech in Asia/Pacific 2018, is the leading Conversational AI Platform for BSFIs to deliver conversational banking services that help banks redefine their digital strategy. With NLP, NLU and Machine Intelligence, the platform enables natural dialogues through messaging, voice and IOT devices. Founded in 2016, Active.Ai expands its global presence with offices in Singapore, India, Australia and North America. Led by seasoned Fintech entrepreneurs, Ravishankar, Shankar Narayanan and Parikshit Paspulati, with experienced team in AI, Machine Learning, Data Science, Finance, Payments, Banking, Mobile and Design.

Please visit www.active.ai or on twitter@activeaibot

Singapore-based Conversational AI Startup Breaks into US Market with Key Bank Clients

Categories:

Active.Ai kicks off a roadmap for US growth with two new bank clients.

February 27, 2018 (New York, NY)— Singapore-based artificial intelligence startup, Active Intelligence Pte Ltd, made a significant step into the United States by signing two important clients.

An enterprise platform for financial services that helps facilitate intelligent micro-conversations, Active.Ai, raised a total funding of US$11.75m from marquee investors, already seen notable success enabling four of the largest banks in India, along with other top-tier banks, and insurance companies in Malaysia, Singapore, Thailand, and the Philippines. Its financial services have span into other verticals like wealth management and securities trading. On the heels of announcing projects with Axis Bank and CIMB Bank being live, Active.Ai now have their sights set on the US market, and already the 24-month-old company has secured a strong foothold in the Americas.

Elizabeth Duke, SVP of Business Development, will be responsible for driving new business opportunities in the US. Duke shared “I’m excited to be part of a growing team that’s focused and committed to the US market. With significant bank customers across Asia, it’s clear the Active.Ai team is ready to deploy at scale in the Americas. The potential is huge to up-end the level of service banks can deliver to their clients”. Duke brings 30 years of global sales and marketing experience in financial services, including senior-level positions at MasterCard, CSAM, Carta Worldwide, and Network for Electronic Transfers Singapore (NETS). She will play a key role in growing the company’s presence across North America alongside co-founder and CEO, Ravishankar, and co-founder and COO, Shankar Narayanan.

Ravishankar, explains: “We’ve already seen a strong interest in our full-stack solution, and these early wins suggest US financial services are ready to incorporate an innovative full-stack AI solution.” He adds, “As Active.Ai scales and enters new markets like the United States, we continue to hire talented people to drive product innovation and growth—2018 will be an exciting year for banks in the Americas.”

Active.Ai runs on a proprietary artificial intelligence engine called Triniti that enables financial institutions to have a meaningful, intuitive engagement with their customers across multiple apertures, using messaging, voice, and IOT devices. This unique solution—built from the ground up by Active.Ai—includes Machine Learning, Natural Language Processing, and Natural Language Generation. It’s arriving just in time for the industry. Shankar Narayanan explains: “The financial services and insurance industries are evolving quickly to remain relevant to changing customer expectations. Using advanced conversational AI, we see an exciting opportunity for these companies to create a natural dialogue and more meaningful connections with their customers using everyday micro-conversations.

Download PDF

Active.Ai raises $8.25 m Series A funding, led by Vertex Ventures

Categories:

Singapore and Bangalore, India, Nov 13, 2017

Active.Ai, a Singapore Headquartered Fintech platform with an innovation lab in Bengaluru, that delivers conversational banking through artificial intelligence (AI), announced a US$ 8.25 million Series A financing led by Vertex Ventures, Creditease Holdings and Dream Incubator. Existing investors Kalaari and IDG Ventures India will also participate in the round. Ben Mathias, Managing Partner at Vertex Ventures and Vani Kola, Managing Director at Kalaari Capital will join Sanat Rao from IDG Ventures India on the Company’s board of directors. Anju Patwardhan from Creditease will join as a board observer.

Active.Ai’s proprietary AI engine, Triniti enables financial institutions to have a meaningful engagement with their customers in an intuitive natural format over multiple apertures covering messaging, voice and IOT devices. This full-stack solution has been built ground-up and comprises Machine Learning, Natural Language Processing and Natural Language Generation. Keeping in mind the requirements of financial institutions, the company offers flexibility of deployment: on-premise or in the cloud.

Founded in early 2016, the company is working with some of the top-tier Banks and Insurance companies in India, Malaysia, Singapore and North America and is planning to expand into other verticals like Wealth Management and Securities Trading. The company’s vision is to become the AI platform of choice for leading Financial Services companies across the world.

Ravi Shankar, co-founder & CEO of Active Intelligence said:  “I am excited by the future possibilities of AI and how this technology will shape the banking and financial world. There is very strong need for banks and financial institutions to evolve fast and empower customers with the ability to do transactions as part of their habitual daily micro-conversations. With the fresh injection of funds, Active.Ai will scale up and continue to hire talented people for the AI team and focus on building the best enterprise product in the market.”

Ben Mathias from Vertex Ventures said: “I am very excited to partner with the Active.Ai team to redefine the paradigm of customer experience in financial services. The threat of fundamental disruption is very real in the financial services space, and companies such as Active.Ai are making it easy for incumbents to not just remain relevant, but to get ahead in an ever evolving market. At Vertex, we have been strong believers in Artificial intelligence helping solve problems specific to industry verticals, and that has been a key driver of our decision to partner with Active.Ai.”

Vani Kola, Managing Director at Kalaari Capital said “We are very excited to have Vertex, Creditease and Dream Incubator join us in this exciting journey of building a world class AI product from India. The fresh round of funding will help us invest even more into technology and expand our footprint into other geographies.”

“Active.Ai’s NLP platform for conversational banking is rapidly gaining acceptance with key banking customers globally.  We continue to be impressed with the passion and capabilities of the team, and are excited to be part of the next phase of their journey.” said Sanat Rao, Partner, IDG Ventures India

Download PDF

Forbes India – 6 July 2018

Categories:

Artificial Intelligence + Most Innovative Companies

Since 2007, when Apple launched its first iPhone, everything we knew changed drastically, especially our behaviour in how we consume the world. With just a touch of a finger, the world, or rather, your world is accessible from your palm.

Fast-forward to over ten years after this significant evolution in the world that changed everything, even the financial services migrated from brick and mortar to mobile. The founders of Active.Ai learnt that, the word ‘unreal’ belongs to the vocabulary of yesteryears, and chose to muse the jargons of tomorrow. With significant advances in cloud computing, Artificial Intelligence, telecommunications and smart devices, it is rather a clue for us that industries that touch human lives are getting disrupted by breathtaking technological advances.

Recognising that internet majors, equipped with higher computing capabilities, financial resources, AI leadership and most importantly, engaged customers, Active.Ai believe that the virtualisation of Financial Services will happen sooner than anticipated. This naturally means that incumbent banks and insurance companies need to think and unthink, learn and unlearn, about their brick and advisory networks.

 

Source: Forbes India

View Article

Citi GPS Bank Future – March 2018

Categories:

 

Q: Tell us about the genesis of Active.Ai?

We started the company in early 2016. I did a trip to China in mid-2015, looking at how financial services were being delivered, and knew China was making huge strides in using AI and the latest technologies to enable financial services. It was fascinating to see how China got there. By the end of 2015, it was clear that the “mobile first” model for financial services, built from China, was going to be disruptive and take on the world.

View Article

The Best 50 Cities for a Startup in the World

Categories:

Source: Valuer.Ai

Singapore has a lot to offer, from modern infrastructure and an educated workforce to a strategic proximity to many emerging and developed markets. Outranking Silicon Valley as the indisputable top and startup capital, which may come as a surprise.

Let’s look back into its history.

Remember the recent competition between Singapore and Hong Kong? Singapore won by earning recognition among investors as the best place for expanding or starting a new business in Asia.

The two are former British colonies, with well-structured governments, free-port trade to foreign investors.
Both depend on economic growth.
The two have a city megapolis and reach between 5-8 million citizens.
Both expanded after World War II, surpassing even Japan’s GDP per capita.
Today Singapore has one of the best education systems in the world and has become a startup hub for entrepreneurs, tech companies and investors.

Here is a list of promising startups in the area: CoinPip, Datarama, Greyloft, Honestbee, Mighty Bear Games, Nugit, oBike, ShopBack, Spark Systems, Chope, 99.co, Active.Ai

View Article

Finacleconnect – March 2018

Categories:

Shankar Narayanan: I think blockchain will definitely be one of the key technologies that have a significant
impact on banking this year. In addition to areas like trade finance and remittances where there is a lot of traction already, I think blockchain will also make a difference in the know-your-customer process bringing in a lot of efficiency and transparency. Secondly, regulations or not, banks will focus on setting up open banking framework to play a larger role in the banking ecosystem. This will be important for banks to develop a new revenue stream and increase their reach through third parties. Thirdly, Artificial Intelligence cannot be ignored any more in banking. 2018 will see more of conversational banking services through chat and voice based interfaces across their digital channels.

View Article

CIO India Review – March 2018 Issue

Categories:

As we gaze far into the year, we can see that the conversational AI landscape is primed for increased consumer adoption. In fact, in a recent survey, nine out of 10 people said they prefer messaging directly with a brand. This year, Apple, Facebook, Google, and Amazon, all leaned into messaging and conversation. While chatbots are still at a nascent stage in the banking industry, bots will quickly gain in sophistication to the point that they will be able to perform all tasks previously owned by customer service representatives. Active.ai, a Singapore based Fintech startup with an innovation lab in Bengaluru, is using artificial intelligence (AI) to deliver Conversational banking services. The startup helps banks redefine their digital strategy for the future, bringing in automation and insightful customer engagement. Built for banking technology, their conversational AI uses advanced natural language processing (NLP), natural language understanding (NLU) and machine intelligence to enable customers to have natural dialogues over messaging, voice or IOT devices.

View Article

Cognitive collaboration: The emerging role of AI in banking

Categories:

Source: CUInsights

What makes a consumers’ experience exceptional in financial services? The consumer wants every transaction to offer simplified authentication, absolute convenience, and personalized relevance. They also expect superfast transactions and for your financial institution (FI) to leverage their transaction insight to do even more.

To deliver on the promise of an exceptional experience the FI has to address three broad areas:

1. Directly collect relevant transactional data from multiple sources

2. Normalize, analyze, and decipher this data in real-time

3. Convert the deciphered data into information that is actionable

This seems easier said than done because of the sheer magnitude of data being collected. And at times even simple transactions can turn into arduous tasks.

Let me highlight this with an example. I received a letter from my airline telling me that my personal information may have been compromised as a result of a data breach. They assured me that they would do their best to assist me and as a token of goodwill they even offered me two years of “identity monitoring and protection.” All I had to do was to fill out a form online.

As I filled out my personal information into the online form I grew wary of the authenticity of the letter from the airline so I called the airline to make sure the letter was legitimate. It took the representative almost 20 minutes to validate things – from finding the actual letter and then reading the letter to explain the content back to me.

At around the same time, I received a similar notification from my credit card provider. I went online and used voice and text to communicate with the brand. It took me less than three minutes to get the same level of confirmation that took 20 minutes with the airline. And, towards the end of my chat session, I even received a call from a representative to make sure that I was satisfied with the answers.

Many voice and text based chatbots are driven by rules-based engines that are limited in their ability to answer conversational questions. Two of my financial institutions offer me voice and text-based banking services (very rudimentary) but I need to know how to phrase/ask the questions.

The above examples are only the beginning. Emerging solutions are powered by an Artificial Intelligence stack (AI programs) that can deliver seamless and useful experiences. AI can power websites, systems, and especially consumer owned devices (mobile phones, Alexa, Google Home, etc.) to do so much more.

Think about these every day and emerging scenarios:

  • Your vacation takes your family to a new location far away from home. As unfamiliar as you are with the surroundings you commute like a local. The perfect route prescribed through filters – road conditions, weather, and your personal preferences.
  • You summon a vehicle that comes to pick you up and drives away after dropping you off to your destination – you don’t even have to say thank you. There was no driver!
  • A robot delivers pizza to your home, the pizza was produced by another robot, while the vehicle with a full kitchen was driven without a driver. Your dinner, all the way from order to fulfillment was handled my machines.
  • You can visit locations through immersive virtual experiences. Realtors are now using this technology to showcase homes without the buyer even leaving their home.
  • Medical providers are conducting “intelligent” pilots where a medical device monitors a patient’s health and provides real-time suggestions to both physicians and the patient.
  • Food distributors can manage their supply chain by viewing satellite data, predicting weather patterns, and tracking crop yields without even visiting their supplying farms.  
  • 90% of the staff at a hotel chain are robots. They interact with guests, perform most tasks, and ensure measurable performance. The robots speak multiple languages and are extremely perceptive to the guests that they serve.
  • The consumer uses voice commands to direct digitally connected appliances for a variety of tasks. From playing music, to understanding weather, to even directing financial transactions – they use their smart phone, or other digital devices.
  • Artificial intelligence can be strengthened by big data and big machines – today, supercomputers are making available transaction intelligence to brands at a fraction of the cost in real time and can even predict likely outcomes.

Artificial Intelligence is slowly becoming the top tier of the technological ecosystem as we know it. A once unforeseeable technology, AI has made its way through the grassy depths, now emerging as a top predator among other existing technologies. Operating as an intelligent learning system, AI has the ability to learn and grow- using data as an input, to create new and smarter outputs. The financial institution must use AI to make the conversion from blue-collar automation to white-collar automation.

Whether it is driverless cars or a robotic concierge, Artificial Intelligence demonstrates its chameleon ability to implement innovative machinery and also, machinery with a touch of personalization. Financial institutions must face the impending threat of AI on the services they offer, in place of its top predator, payment institutions. Capitalizing on the “sixth sense” of AI, empathy, financial institutions can focus on seven key areas of focus where Artificial Intelligence can be applied:

1. Perfecting Service: Recognizing and treating people well
2. Circumventing Fraud: Protecting people – information, transactions, and data
3. Precise Offers: Serving the consumers what they need, when they need it
4. Optimizing Risk: Dynamic models to balance between success and failure
5. Regulatory Compliance: To ensure that you are doing what you are required to
6. Intuitive Research: Having anytime, easy access to information
7. Empathetic Employees: Allowing employees the time to listen, connect, and serve

These topics harbor the idea of secure, superfast, and personalized transactions, three characteristics that compete with a mounting AI presence. The financial institution is capable of coupling precision with personalization in order to stay in the game.

In a world where artificial is more real and mainstream, new and more intelligent competition has emerged. This competition has the ability to constantly evolve, creating consistently better results, and keeping it ahead of the curb. We should capitalize on these intelligent machines by seamlessly integrating artificiality into our own services, changing the connotation of AI to an environment of cognitive collaboration that is defining the future of financial services.

Personalized Connected Experiences

Read on CUInsight

Transforming the Business Banking Experience with AI

Categories:

Source: Medium

As many of the industry analysis reports from McKinsey and BCG over the last few years point out, corporate banks have been the last to join the digital bandwagon. Corporate internet and mobile banking took off to some extent over the last decade. Straight through processing (STP) for various transactions through direct integrations between the ERPs and the banking systems has eased out a lot of friction in the business banking processes. However, the overall digitalisation is way less than the other lines of businesses in banking.

Even though many areas of corporate banking have undergone some amount of digital transformation, when it comes to customer experience, there is still a lot of dependence on the RM, even for simple queries starting from balance checks to financial transaction status. Such engagements serve the immediate requirement of the client but are of no real “value added” from a costly channel like the RM.

These engagements extend to contact centers through calls and emails which are all extremely high cost and high turn-around-time channels. The interactions are not exactly delightful when the client has to either wait in a call queue or be kept on hold for the agent to get the appropriate details to answer the query.

Most RMs engage with their clients through calls and other messaging channels which are outside the institutional purview. Thus, these interactions are not recorded in any way. Adding to that is the security consideration of such exchanges. The only form of authentication is the RM’s belief that he or she is speaking to the actual client on the other end of the channel. It is also well known that the attrition in the RM community is very high and along with the RM, all the insights on the clients, which should have been added to the institutional knowledge, also get lost to the institution.

Now cut to retail banking which has been at the forefront of digitalization. While “mobile first” was the mantra a few years back, today it is “AI first”. As various reports point out, AI investments in banks are the highest in the area of customer experience. We are seeing a big movement in retail banking customer experience where the interaction is moving from structured, form-based channels like internet banking and mobile applications, to unstructured and natural-language-based conversational engagements. There are multiple channels that the banks are opening up, ranging from chatbots on their websites and mobile applications, to going where the customers are, like Facebook messenger, WhatsApp and so on, to voice channels like Alexa and Google Assistant.

As mentioned earlier, corporate banks were always more into unstructured, natural language interactions. However, through these high cost channels and at times, channels that the institution has no insight into and with human authentication only, it ends up with no records. Moving to another conversational channel, which can be even lower cost than mobile and internet banking and available 24/7, can be the next big digital transformation for banks to serve their enterprise clients.

With AI based conversational banking, the banks can reach out to the business clients in a channel which they are already using. Most of them use the channel for their internal collaboration or to interact with their RMs, which means that the institutional data and knowledge will be saved. This will be in addition to internet banking and mobile banking application today. However, over a period of time, the conversational channels can see a much higher adoption. The AI based conversational channel should have the ability to seamlessly handover to the RM or a human agent when it cannot cater to the customer request, to ensure the service levels remain at a high level even in the early days when the system has had little training. In fact, the best experiences can be provided by such human-machine combinations instead of any one of them.

Conversational channels can be deployed as extensions of the website, internet banking and mobile banking, through additional channels like Skype, Lync, to name a few, that businesses use for their internal collaborations. Banking through Skype, Lync type channels by business clients essentially mean that the corporate users don’t have to open a different website or mobile application but instead continue with their banking from where they are collaborating with their colleagues for other purposes.

Besides these messenger channels, for immediate, short and critical queries of the CXOs, banks can create conversational experiences through voice channels like Alexa, GA and so on.

Conversational channels in business banking are relatively unexplored areas. On that note, it has great potential only if the channel specific customer experience (CX) design is completely reimagined for the channel and not replicated from web or mobile. Even the voice interface design (VUIs) should be differently thought of from the messenger conversations. Only then will the adoptions of these channels soar.

The conversation experience design should be a combination of the target persona in the corporate and the channel and must start from the requirements of the persona. A Juniper report shows that there is an average of four minutes savings per call when transactions are done over chatbot. In addition to that, a Medici report shows that 70% of customers today prefer a messenger channel over a call. Our belief is that a well-designed conversation engagement through any channel should bring down the time further and provide a delightful experience. The entire experience design should be on an extremely strong backbone of conversational AI, whose (1) natural language understanding (NLU) should be able to understand the nuances of complex human conversation and the corporate banking domain ontology, and (2) Machine Learning capabilities keep making the conversations more individualized to the corporate and the particular user.

One of the possible ways to start off the conversational channels for business banking is to analyze the high-volume interactions in the contact centers and with the RMs. While the contact center data will be easier to assimilate, the RM data points are equally important since that is probably the area that needs to be brought into the institutional purview asap. Such analysis will not only help the banks prioritize the type of interactions to provide in this channel, but the actual conversation data can be used for training the AI system on the utterances of the customers and the possible replies the bot can provide.

Providing a conversational channel for prospects (who are not yet customers of the bank) could be the very first step for business banking. This could be a completely new channel on the bank’s website to help out the prospects with products that they are seeking or providing the simplest and fastest way to frequently asked questions (FAQ). The experience can be truly delightful only if the AI engine understands the queries exactly and the replies can be given to the point with minimal redirection to elaborate documents. Wherever the queries are not clearly understood, a seamless and quick handover to a live agent could help keep the experience frictionless. This combination of AI based conversational channel and live agent, only in extreme cases, could be a great recipe for client conversion instead of an abandonment.

FAQs could be a good start for the conversational channel for existing customers as well. A lot of their generic queries do land up with the RMs and call centers. Having a simple self-help in their corporate messenger could be an immense help. The next phase could be to take up service requests, such as statements and certificates, and transaction approvals, like corporate multi-user approvals for high value transactions, through this channel. High volume financial inquiries like remittance status, beneficiary credit, invoice payment stage and so on could be the subsequent phase of the conversation channel. Based on the adoption, the conversation channel could provide all the high-volume interactions over a period of time.

Banks can also look at virtual assistants as a part of business internet banking that could be completely conversational. Corporate internet banking with its multitude of options could overwhelm the user and drive her to the call center. The conversational assistants can be invoked within the internet banking session that can help her navigate through the system or help her with a balance inquiry when she is in the middle of approving an ad-hoc payment without having to move out of the approval option. This can prevent that costly call to the bank and save her a lot of time as well.

In summary, we see that conversational channels based on powerful natural language capabilities and continuous machine learning can help banks provide their business customers a delightful, always available experience and reduce the usage of costly and cumbersome channels. That will essentially mean that they reach out to RMs and call centers for real value adds like advice on new facilities or better deployment of funds that can be more revenue generating for the bank… and a fringe benefit could be higher retention of RMs.

To know more on how to transform banking experience, please visit Active.Ai.

Source of References: BCGMcKinseyGoMedici

Read on Medium

Staying Sane and Optimistic amid the AI Ballyhoo!

Categories:

Source: Medium

At NIPS 2016, there was an unprecedented story building up. Something that got every AI enthusiast agog about an unknown AI startup ‘Rocket AI’.

The names associated with the hot startup were pioneers in the AI field and it was informed to the media that there was major announcement soon to come. There was even a workshop held, where one of the researchers explained about the concept of Temporal Recurrent Optimal Learning to the house full of researchers and media personnel.

The whole community was abuzz with the jargons “Temporally Recurrent Optimal Learning”, “Jacobian Optimized Kernel Expansion”, “Fully Automatic Kernel Expansion”, a couple of which were coined by a leading AI researcher. These made rounds on web with a hype so strong that it got 5 Major VCs reaching out to them for investment.

There were rumours about Rocket AI’s acquisition as well — all within a day.

Only to be figured out, later, to be a joke!

Temporally Recurrent Optimal Learning = TROL
Jacobian Optimized Kernel Expansion = JOKE
Fully Automatic Kernel Expansion = FAKE

If you still don’t get the joke, you should probably get yourself the highly coveted RocketAI T-shirt @$22.99 a piece.

 

The Rocket AI launch story at NIPS16 was a great lesson on why it is imperative to disentangle noise/hype from the real advancements around a technology on surge.

But let’s be honest, can we inculpate anyone when they fall for the AI ballyhoo? Such hype has always been associated to technologies with potential to change the world. In many ways, it already is. Amidst the extravagant hype, there has been multitudinous success stories as well. Autonomous vehicles, AlphaGo victory over world №1 Go player, OpenAI DOTA2 bot challenging top professional gamers, surge in successful applications in healthcare & fashion owing to advancements in computer vision, is a testimony to the fact that AI has arrived and it’s here to stay.

Of late, there has been a myriad of NLU/NLI/QA datasets crowdsourced and released for the research community. It has undoubtedly propelled the DL research efforts for NLP applications. BABI inspired multiple forms of memory networks, SQUAD/newsQA impelled BiDAF, AOA, mnemonic reader, fusion-net etc, SNLI/MultiNLI,RTE exhorted multitude of attention networks.

However, an agonising trend that concerns me is the over-reliance of young AI practitioners on SOTA (State-of-the-Art) for identical problems. Whether you’re a researcher involved in primary research on AI, a product developer utilising the existing frameworks/algorithms to a given business problem, or an executive evaluating a product, one should always be mindful of the capabilities and limitations of any such SOTA, and its applicability to the given use-case/dataset.

After all, “There are no free lunches in AI”.

Lets have a look at a few observations around the issues with a few popular Datasets used for benchmarking NLU/NLI algorithms and the State of art solutions for them.

 

Issues with the popular State-of-the-Art solutions and the datasets

1. Squad performance State-of-the-Art models dropped drastically after adversarial examples were added in V2.0 compared to V1.1.

2. Best performing model on Squad dataset only yields close to 40% f1 Score on NewsQA testSet.

3. InferSent seems to rely on word level heuristics for high performance, for SNLI challenge. Most of the contradictory sentences have no overlap in words and with high overlaps, it is more likely to be entailed. This indicates InferSent may be simply learning the heuristic that high overlap could mean ‘entailment’, presence of negation could mean ‘contradiction’. For pairs with high overlap between words with the only token difference among the pair being an antonym, they are classified as ‘contradictions’. Presence of an additional non-antonym word throws the predictions off, classifying it as ‘entailment’.

4. Try training a fastText classifier on SNLI dataset and ensure that the model doesn’t get to see the premise but only observes and predicts based on the hypothesis. Be ready to be baffled, because the prediction accuracy would be way higher than some of the baselines! A task which makes no sense to be solved without considering both hypothesis and premise together, it’s quite amusing to find a model doing well without the premise.

5. Quantifying determiners and superlatives were mostly included to make the sentences look similar but always present a deviation from the hypothesis. Thus, a ‘contradiction’ in most of the cases.

6. Negation, presence of hypernym and hyponym, and overlap of words largely contributed to the entailment class.

These unprecedented benchmarks (SNLI inference problem: train accuracy 95%, test accuracy 91%) might give an illusion that natural language inference is already a solved problem. However, these evidences say otherwise.

Tesla AI director Andrej Karpathy showed with few simple examples of how a deep-learning model could be fooled, by adding little amount of noise.

Panda gets recognised as gibbon with high confidence.

The Graph above shows the computation requirements of different AI solutions against their timeline of existence. Is it just the compute power & more data or better algorithms?

Coursera co-founder & Google brain founding member Andrew NG claimed the AI systems could diagnose diseases from X-ray reports way better than radiologists.

The results later proved AI still has a long way to go, in making radiologists totally obsolete.

 

 

Some interesting excerpts from The former Facebook AI head Yann Lecunn’s interview with Spectrum.

 

Research frontiers

Lets look at some of the recent AI breakthroughs in computer vision and NLU. Nvidia, OpenAI, Microsoft and Google have been the forerunners.

Below are the image outputs from the OpenAI model which performs Mix-match and Image manipulation. The output is extra-ordinarily high definition and seems natural, although they aren’t real human beings. The third image shows how Nvidia’s model re-creates an unrecognisable object in an extremely blur image automatically.

 

Source: https://blog.openai.com/glow/ (Image manipulation and Mix-match)

 

Source: https://www.dpreview.com/news/0229957644/nvidia-researchers-ai-grain-noise-images-photos

 

Learning with lesser data: Meta learning Meta learning is the process of learning to learn. A meta learning considers a distribution of tasks, and generalises to learn a task efficiently with a very less amount of data. Reptile, MAML are a few popular Meta learning algorithms. Most of the Machine learning usecases at present, are dependent upon supervised approach of Learning, requires a lot of annotated data which is scarce to find. Breakthrough in meta learning approaches can really pave the way to building learning systems which can learn with the very less amount of data, just like how humans learn with substantially small amount of observations.

Automated annotation: There has been a lot of progress in this front, with the evolution of powerful sampling strategies and prediction reliability estimates which can help in sampling a minimum set of instances required to be annotated for the model to approximate a function good enough to replicate the behaviour of the model trained on the full dataset. These frameworks can be applied to automate the annotation process by reducing the manual efforts to a great extent.

Multitask learning: Can you cast each of these NLP tasks as question-answering task and solve it with one network? Question answering, machine translation, summarisation, natural language inference, sentiment analysis, semantic role labelling, zero-shot relation extraction, goal-oriented dialogue, semantic parsing, and common sense pronoun resolution?

Evaluation framework to identify annotation biases: The issues of annotation bias is very much evident by the issues identified in some of the well known public corpuses used for research purpose (SQUAD 1.1, SNLI ). A lot of the existing State-of-the-Art algorithms may not perform high on these datasets if these annotation biases are removed, which will only entice the research community to push the limits and explore better approaches to tackle these problems.

Unsupervised approaches of learning: There has been a lot of research in supervised learning approaches, and its not an unknown fact that given a sizeable training data, it’s no more difficult to train a system to do a varied activities of prediction/recognition/generation etc. The data gathering, annotation, curation is an expansive process and for a lot of real-world problems this becomes a bottleneck.

Unsupervised Sentiment Neuron, Unsupervised Language Modelling using transformers and Unsupervised Pre-training, yielded encouraging results.

Contextual and efficient word representations. Word2vec , Glove & Cove were a good start. FastText made it better, ELmo seems to have surpassed the others based on its performance over a lot of NLU/NLI tasks. The contextual word representation for the words have proved to be quite helpful in downstream tasks.

We are starting to see impressive results in natural language processing with Deep Learning augmented with a memory module. These systems are based on the idea of representing words and sentences with continuous vectors, transforming these vectors through layers of a deep architecture, and storing them in a kind of associative memory. This works very well for question-answering and for language translation

Search for Better optimisers: Using machine learning to find better optimisers proved to be quite beneficial. Optimisers play a pivotal role in the performance of Deep learning architectures. Some of the most commonly used optimisers being Adam, AdaGrad, SGD etc. The search for efficient optimisers is just as important as discovering better optimisation algorithms. Google AI showed that reinforcement learning can be utilised for searching better optimisers for deep learning architectures. PowerSign and AdaSign were discovered using the same, and have proved to quite efficient for many DL usecases.

Zero-shot learning/One-shot learning: Can a supervised learning model be trained to predict a class that is not present or are entirely removed from the training data?

 

Interpretation vs Accuracy

An interesting debate surfaced up at NIPS 2017, with Ali Rahimi and Yann LeCun locking horns over Rahimi’s remarks of Machine learning becoming the ‘alchemy’. Blaming the present Deep learning systems to be a black box or at best an experimental science, Ali maintained the AI systems need to be based on verifiable, rigorous, thorough knowledge, and not on alchemy. LeCun argued that the “lack of clear explanations does not affect the ability of deep learning to solve problems!” While LRP(layer wise relevance propagation), deep Taylor series and LIME (Local Interpretable Model-Agnostic Explanations) are a few methods being utilised to make these systems explain their predictions, it is still in its nascent stage. Or do we need to settle this debate by saying, “It might just be part of the nature of intelligence that only some part of it is exposed to rational explanation. Some of it is just instinctual or subconscious, or inscrutable.”

The voice in favour of supposedly imminent AI winter is just as strong as the ardent optimists who believe AGI is just around the corner. Although the later may not be likely anytime soon, The former is more likely if AI breakthroughs are overhyped beyond truth.

“AI researchers, down in the trenches, have to strike a delicate balance: Be optimistic about what you can achieve, but don’t oversell what you can do. Point out how difficult your job is, but don’t make it sound hopeless. You need to be honest with your funders, sponsors, and employers, with your peers and colleagues, with the public, and with yourself. It is difficult when there is a lot of uncertainty about future progress, and when less honest or more self-deluded people make wild claims of future success. That’s why we don’t like hype: It is made by people who are either dishonest or self-deluded, and makes the life of serious and honest scientists considerably more difficult.” Yann Lecunn

 

Ashish is the VP of AI Product for Active.Ai. Find out what the future holds for Conversational AI at www.active.ai

Read on Medium

Conversation Design and breaking from shackles of App mindset

Categories:

Source: Medium

Conversational Interfaces are relatively new. Facebook was one of the earliest to popularise it and this was a little over 2 years ago. The last 4 decades have seen human computer interfaces stabilise around point-and-click GUIs evolving all the way from Apple Macintosh introduced in 1982 to the iPhone (that introduced multitouch and swipes to the masses) launch in 2007. Massive leaps has been made in those 4 decades in defining usability standards and guidelines for all platforms like desktop, web and mobile applications. Conversational interfaces require a whole new mindset to realise its potential to the fullest.

As you read the rest of this article, you will find that a lot of examples are provided in banking context, which is where most of our experience and audience lies. But I see no reason why most of these principles cannot be applied to pizza ordering, for example.

Application interfaces had one requirement that drives its design and all usability guidelines around it. If you make an application module, for example, to do a bill payment, the best way to manage this is to automatically present the options on screen. Even those requirement settings that rarely changed, like for example, the source account for making payment, or the payment date set as default to a certain day and time, or even optional elements, like comment.

What defines the application design is the fact that, unless the options were presented on screen, it cannot be used. Users have also developed a sense of selective blindness overtime as they got more familiar with the interfaces. So they come online, just select the biller, enter the amount, just glance if the other default fields are correct and complete the transaction.

Moving to conversations. The first instinct of an experience designer is to look at current design of mobile/internet banking apps and design the conversation based on that. That is one of the most terrible way to start. Once you have seen the mobile/internet banking screens, it is very difficult to un-see it. We have some cases where a complete form in an internet banking experience is replicated into the conversation.

This is what we say to all our customers. When designing a conversation for any given use case, close your eyes, imagine your smart and intelligent secretary/spouse just walked into the room and you asked him/her to remit the rent. He/she already knows who your landlord is and how much the rent is. The response you would expect is “Sure, I will do that right away. Can I get you a coffee or anything else?”

Now imagine he/she asks “To whom do you want to pay? How much? From which account? When?” You will probably look for another secretary. If he/she is your spouse, too bad. 🙂

The point is, your bank, unlike most organisations on this planet (save for Google and Facebook), knows you the most. Replicating that secretary scenario on a conversational channel is not even AI, it just requires simple historical look up against users’ past transactions. Yet, so few are thinking along such lines.

The guideline for a successful conversation is this. If the user cannot do it faster on a conversational channel compare to a mobile application, they wont don’t do it again. I transfer funds regularly to my wife for our expenses and measured the time it take to complete the activity on the conversation platform. It is definitely faster.

1. Go to bank.com
2. Open the bot
3. Type “Transfer 50000 to wife”
4. Authenticate
5. Confirm
6. Enter One-Time password (OTP) received via SMS
7. Done.

This is nearly perfect for a conversational channel. One small change I would make is to avoid the confirmation step if this is a regular pattern of the customer and jump directly to “Ok. I am transferring 5000 to your wife. Please enter the OTP to proceed.” If this is on another registered channel like Facebook Messenger, the authentication step can be skipped as well.

The idea is to design conversation for pure audio channel from the start. That keeps you grounded. Facebook did a great job in defining standard templates. The purpose of those templates such as cards, lists and carousels is to provide information and avoid typing as much as possible. Opening a design tool like Sketch/Balsamiq to design the conversation flow for a usecase is a sure shot way to derail the thought process and may end up too app-like.

Conversation designer should be more like a playwright than a UX designer. Once the playwriting process is complete, you can then add lists and cards at appropriate places within the flow if the channels allow it.

Another key aspect of designing good conversational applications is the 80–20 rule and ability for users to change the assumptions. I will illustrate via an example below from a recent experience.

Recently when we designed a flow for opening a term deposit with a bank, one thing lead to another and the proposed flow ended up with 15 steps that was replicated based on its other channels.

It goes like this:

1. How much you want to deposit
2. One time / Recurring
3. What Tenure
4. What happens on maturity… and so on.

One of our co-founders threw a challenge. Why can’t we do it in one step?

And this is how we did it:

1. What is the information that we really need from the user? It is just the amount that he wants to deposit.

2.Now the bank knows that 80% of the users who do recurring deposit, the amount is less than INR20,000. Therefore, it is quite fair to assume recurring or single deposit based on that amount.

3. Now, the bank also knows that for a recurring deposit, what the most common tenure chosen by 80% of the users will be or if it is one time, the bank can assume the tenure is to get the best interest rate possible.

4. The bank also knows what the behaviour will be for the majority upon maturity.

So, essentially, the flow goes like this.

User: Open a deposit for 50k

Bank: Sure. The best I can offer you is 7.6% for a 11 month term. Proceeds will be credited back into your account after 11 months. Interest of INR 767 be credited quarterly into your account . Shall I proceed?

User: Yes, Please.

In Summary:

Design the conversation so that it is extremely fast for 80% of the users. Its not fair dragging the conversation just to cater for the other 20% outliers.

Now comes the beauty of conversational channel. Although the flow was designed with the 80% in mind, it does not mean we are ignoring the remaining 20%.

As an example, those 20% of users can simply ask “Can you make the tenure 5 months please?” and customise any assumption that the system made for them. And over time, if a user always tends to change his tenure to 5 months, the conversation system should start assuming 5 months tenure for that particular user. Or if the user has no fixed pattern of deciding tenure, it makes sense for the bot to ask for the tenure every time.

This is true personalisation. That is what your secretary/spouse would do. This allows the conversation system to be extremely fast for the 80% on day 1 going live. And it evolves automatically for the remaining 20% to be fast for them too, based on their patterns.

Before signing off, I would like to highlight that not every usecase fits well into a conversational interface. Apps and websites will not fade away. Try not to fit every usecase into a conversational journey. Any usecase that is too long (like opening an account and begin a brand new relationship with a bank, or apply for a home mortgage) are not naturally suited for conversations. Although conversations is the in thing now, users are better off filling a form on a website or mobile app where he can go back and forth, changing inputs etc.

Conversational interfaces provide a very powerful way to interact with your users. Design the system that can leverage the knowledge you have about your customer base and create journeys that wow them.

Read on Medium

Triniti meets Oracle

Categories:

I’ve been working with Active.Ai for more than a year now, leading operations and product for the North American market. I focus on identifying product requirements unique to the target market, working with our clients, prospects, analysts and gathering market research information.

Active.Ai’s key difference is our team – I couldn’t ask for a more passionate, talented, helpful and smart bunch of colleagues. Together, we have built a platform, Triniti, that’s at least a year or two ahead of the market place in terms of allowing a natural conversation with an AI system. Our focus on Financial Institutions and their needs has paid off as we have produced pre-built user journeys and datasets that accelerate project development and deployment.

When Oracle Banking team invited us to join them for the Industry Connect event in New York City, we jumped at the opportunity as we share the same ethos towards creating a secure and scalable banking platform. We believe that we complement Oracle’s core banking systems with the front end NLP capabilities for next-generation user interfaces.

For a demo day like this, we prepare by profiling the audience – is it more technical or more business oriented? Based on this, we draw up a presentation deck that addresses the relevant features and benefits of our platform. We include existing collaterals, and if time allows, we include custom demo materials. We made sure that every bit of Triniti’s capabilities in delivering top conversational experience is showcased.

At Industry Connect, Oracle provided us with an opportunity to work together with them in a ‘hackathon’ format. This was hectic but fun, and the end result is there to see in the video – a conversational website utilizing Oracle Banking APIs.

Madhav Mehra is the VP Product & Operations for Active.Ai for the North American market. Find out more on how you can ride along their journey into the future of Conversational AI at www.active.ai

How do we anticipate the future?

Categories:

The last few years have seen significant advances in cloud computing, AI, telecommunications and smart devices. Every industry which touches human lives is getting disrupted for the better by technological advances at a breathtaking pace.

At Active.Ai, our team has been imagining the future of financial services at our skunk works style innovation hubs.

This helps us bring ideas from experimentation to proof of concepts to pilots at great pace.

I recently had a peep into our own Matrix for wealth management, where I walked around a mocked up virtual portfolio, pinched and moved graphs across asset classes, wearing a Holo Lens. It felt like so much like science fiction, that I had to pinch myself!

We can very well imagine a future, where an Advisor and Client are engaged over conversations, on such platforms. Banks can create virtual branches and tellers. Insurance companies can be engaging in virtual claims discussions. Robo-advisors can add personas.

Yes, it seems unreal but we can see that in near future, these capabilities may become mainstream in customer engagement. 2007 was the launch of the smartphone (Apple’s iPhone). In a decade, everything changed and financial services moved from Brick and Mortar to mobile.

Given that Internet majors, having tremendous computing capabilities, financial resources, AI leadership and most importantly engaged customers, we believe that the Financial Services virtualization will happen sooner than anticipated. Incumbent banks and Insurance companies will need to think (or rather, ‘unthink’) about their brick and advisor networks.

Maybe in the year 2021 you would be greeting your banker wearing smartglasses in a virtual branch.

Do you see this possible in the future? Tell us what you think: hello@active.ai

 

Source: Medium

How Chatbots in Financial Services are Evolving with NLP and NLU

Categories:

AI-powered chatbots can understand your specific banking needs using NLU — and also detect how you feel and predict what you’ll say next.

Processing human language is a challenge for artificial intelligence — the way we converse is nonlinear, irregular, emotional, and full of context. The goal is for AI to hold a back-and-forth conversation with a human in a way that feels natural — despite the fact that it’s with a machine. Natural language processing — or NLP — is the first step to making this a reality.

In customer service-focused industries, like banking, chatbots equipped with NLP can analyze, process, and communicate with users, using language they understand. NLP techniques categorize customer data by tagging parts of speech, correcting spelling, and re-formatting numbers and dates into something the machine can read.Read more…

Digital Davids vs Financial Goliaths

Categories:

Active.Ai Ravi Digital Davids vs Financial Goliaths
We live in an interesting time of disruptions. Like many industries, the financial services sector is facing its Kodak moment. The incumbents have the experience, the challengers are creating the experience. Here are a few of the many moments that I feel will define the next decade for the incumbents.

Read more

Contact Us






Request a Demo