How Artificial Intelligence will transform Banking


Circa 2030. Somewhere in Mumbai…

As Adam wakes up on Monday morning, he knows today is a special day. Adam has been training for this for 2 months now. He finally has access to Subbu’s account and before he even realises, Adam would have made the transfers.

An hour later while Subbu is on his way to the office in a driverless Tesla, listening to old Daft Punk classics, his Apple watch beeps. He scrolls through each notification and realises that Adam has already transferred ₹600,000 from his account. But Subbu isn’t worried.

Adam, his Personal Assistant bot, finished the mandated training of 60 days just yesterday and can now make bank transactions on his behalf – initiate remittances and investments, make utility and card payments and book tickets and hotels for the next family trip to Australia.

Times, they are changing. Again.

The scenario in 2030 may seem like fiction but the speed at which Artificial Intelligence (AI) is evolving, this may be a very simple use case. If you are paranoid about AI taking over the world, like Skynet in Terminator, a  report by Stanford University titled ‘Artificial Intelligence and Life in 2030’ allays some fears:

Unlike in the movies, there is no race of superhuman robots on the horizon or probably even possible. And while the potential to abuse AI technologies must be acknowledged and addressed, their greater potential is, among other things, to make driving safer, help children learn, and extend and enhance people’s lives.”

While AI technology may still be some distance from reaching a high level of ‘consciousness’ to ‘manipulate’ a coup, that doesn’t mean it comes as a clean package.

Let’s get over with the bad news first

Banks generate and consume a large amount of data. AI relishes data. Banks have standard and repetitive processes. AI gobbles up standardisation. The problem is conspicuous:

  1. Repetitive jobs in back-office operations like screening standardised forms and processing routine transactions would be automated.
  2. Contact centers and IVRs would no longer be dumb and you needn’t have to wait for a human John or Arvind to answer. John and Arvind would need to look for a new job.
  3. Parameterised retail products like credit cards (in whatever form) and small ticket loans are fit cases for automation. Analyst jobs, Risk and Credit Middle offices may all be automated and the lone human supervisor handling these functions may be more techie than a bankie!
  4. And Branch staff. What branch??

Job loss is inevitable in banking and many other sectors.

Another risk of AI is the danger of centralisation of power. Stephan Hawking and Elon Musk have been vocal about responsible use of AI. Would those who have access and control of AI, become czars of the new world order?

If not already worrying about it, Government think-tanks should start mulling over the socio-economic impact of AI, in not so distant future.

There are no straight answers, but there are some silver linings.

And the Good news

Despite the apprehensions, AI is still a powerful tool to solve long-standing problems of humanity. It can improve access to medicine and diagnostic tools

US leads the investments in AI and its growing in UK, Israel and India.

for the remotest parts of the world. Self-Driving cars can help prevent thousands of accidental deaths and even help with climate control. And yes, it can simplify banking- relieve the pain of redundant forms, cumbersome documents and slow processes and even manage investment portfolios. It can help banks be nimbler and still offer a wider range of services, at the fraction of a today’s cost.

All this culminates into efficient systems, better compliance, improved profitability and happier customers. It is for this reason that the biggest of the corporations are betting themselves on AI. You can see some interesting data here. The effect in India is trickling down with more and more local startups waking up to this opportunity.

AI may just be the golden ticket for the ‘Guilty Banker’ to redeem himself and bring ‘trust’ back in banking.

Here are some use-cases where AI is or will soon, transform banking.

  1. Chatbots: Improved conversational ability

With the improvement in Natural Language processing capability, it would be possible to have meaningful and complex conversations with the Chatbots (computer programs that can interact in natural language with humans).

Chatbots can respond to both speech and written text.

RBS has been experimenting with Luvo, its personal customer service assistant and DBS has launched its Digibank in India, supported by a 24 x7 digital assistant based on AI bot Kai by Kasisto. There are others like IPsoft’s Amelia, which the company claims can “emulate human intelligence making her capable of completely natural interaction with people”, implying it would understand if you are angry on in a tearing hurry and respond accordingly.

Chatbots are available 24×7, in any language and can drastically reduce the cost of customer service infrastructure.

With the development of Apps like Microsoft Translator, that can translate between any two languages, banks wouldn’t need to hire multilingual resources. They may well hire all German speaking executives to handle Japanese customers! By extension, expert knowledge would not be bound by the language barrier and banks may access and offer expert advisory from any corner of the globe.

Ira, is the first Humanoid branch assistant deployed by HDFC Bank

Improvement in Speech recognition and Reinforced learning can potentially change our interactions at bank branches. Empowered with conversational ability, Humanoid Robots would become mainstream branch resources and go from the current stage of saying ‘Namaste’ to handling all Teller or other branch level queries.

  1. Robo-advisors: Manage money better

With commoditisation of routine services like Payments, Customer’s relationship with the bank is becoming purely transactional. The myopic view of profitability has led banks to restrict their advisory services, for whatever it’s worth, to large ticket or HNI portfolios. For most people, the only advisory they ever got from banks is the push to buy third party products like insurance, while ignoring core products or services. It’s surprising that banks couldn’t offer as basic a service as expense management which Intuit’s Mint (and many others) now provide.

Now, with Robo-Advisors (an algorithm based online wealth management service) like Wealthfront and Betterment, the wealth and portfolio management services are becoming accessible, cost effective and democratised. Your investments may soon be managed by an AI Advisor, giving you flexibility and real-time inputs minus the bias of wealth managers who push products to achieve their targets. And all this at a much lesser cost.

While the current Robo-Advisor models are hybrid, requiring significant human intervention at the back-end, these are evolving very quickly to become fully automated.

  1. Better Credit Risk Assessment

The development of tools that move away from statistical rule-based engines to cognitive learning models, it is getting easier to automate parameterised or score-based products like Retail loans and Credit cards. AI empowered systems would soon handle deviations in the parameters, which presently require human intervention.  These bank products are low-hanging fruits for AI implementation and approvals for small ticket lending will soon be real time instead of days.

On the other hand, large ticket lending to Corporates or Small and Medium businesses requires a detailed analysis. The assessing credit officer uses his/her judgment to decipher complex deviation scenarios. For eg., if the financials show a decreasing trend in Tangible Net worth, the officer relates this to other factors like reduced cash profits, diversion of funds, high capex or abnormal increase in current assets. He evaluates these multiple scenarios simultaneously, meets up with the customer and corroborates with other available parameters like Current ratio performance, internal and external ratings and management’s ability to infuse capital, before arriving at the decision. The key skill of the Credit officer lies in determining the materiality of each parameter, going beyond simple correlation and deciding on a more likely future scenario (aka projections).The banks would call this ‘touch and feel’ of the large ticket credit. This approach is fraught with miss-outs, inconsistency and unpredictability.

As AI evolves from job specific Artificial Narrow Intelligence or weak intelligence to more advanced Artificial General Intelligence (AGI), with human-like judgemental abilities, such scenario building would become automated too, bringing much more predictability, better reading of Early-Warning Signals and help reduce delinquencies.

AlphaGo was able to decipher 10^170 permutations in the ancient Chinese game ‘Go’

This increasing ability to handle complexity was demonstrated by AlphaGo, a Google-owned AI software, where deep neural networks and machine learning were used to decipher zillions of permutations and defeat a human champion in a complex Chinese game ‘Go’.

While it is still short of building the ‘intuitive’ judgment, it’s only a question of time before it reaches there.

  1.  Anti-Money Laundering (AML) and Compliance

Currently, banks rely on AML tools like Factiva and Thomson Reuters’ World Check, which are primarily database checks to identify matches. There are many products available that monitor transactions and identify anomalies, based on set rules. The further development in text analytics would make the plethora of unstructured data, spread across the web, being converted to usable information.

AI can help plug the gap in AML leakages

One use case which I believe can help AML initiative is building a common pool of learnings of banks across the globe. Imagine a scenario where it is possible to find out total remittances made by a person, across banks to a sanctioned country like Syria. At present no bank would share an account level information with others, considering the data security and competitive considerations. But with Homomorphic encryption, it may now be possible to build tools to share data in encrypted form and perform functions without the need to decrypt the data. In simpler language, it means it is possible for banks to share the data without revealing what that data contains and it would still be possible for the user or authorities to use the data for required checks.

  1. Regulatory advisory and legal

Banks operate in an increasingly dynamic regulatory environment. The constant fear of being on the wrong side of the law is one reason why bankers have become over-prudent and avoid taking even logical calls. There is a whole ocean of circulars, statutes, laws and guidance that a bank and its customers need to navigate. It’s even more relevant in a developing economy like India where both complexities of transactions and regulator’s response are evolving.
In my career as a banker, there have been numerous occasions where customers seek clarity on FDI (Foreign Direct Investment) or ODI (Overseas Direct Investment) norms, requiring approval, condonation or clarity from RBI (Reserve Bank of India). The moment such a grey area is encountered, the transactions are put on hold till the time regulators respond.

But AI would catch up in these complex use-cases as well. Ross Intelligence is an AI augmented legal advisor that makes it easy to search through legal literature. It is built using IBM Watson and can answer legal queries. Scanning banking laws and regulations may just be a simpler use case.


“By the 2040s, non-biological intelligence will be a billion times more capable than biological intelligence” – Ray Kurzweil

While AI is growing fast, the development is not linear or predictable. It may be difficult to pinpoint which scenarios would unfold first and to what extent – would it be chatbots becoming more human-like and make call centers redundant or Robo-advisors taking over the money management completely or Middle offices becoming automated or credit decisions becoming real time.

Some complex use cases like corporate debt, large ticket lending, investment banking, stressed asset management may have some lead time, but it may be sooner than later before AI catches up. In fact, it already is. Investment bankers may want to see how Numerai, a crowdsourced Hedge Fund model is being built to have better predictions on asset returns or Alpha Sense, a search engine optimised for financial search can bring company information quickly to a corporate banker.

We are very near an AI inflexion point and the effects are not limited to banking.  It will disrupt the way we live lives.

Banks cannot remain mute spectators or have a piecemeal approach to this. They need to have a full-fledged team working on AI with large scale investments. It’s time they have robots on their side.

By Amit Balooni


Leave a Reply Text

Your email address will not be published. Required fields are marked *

error: Pls mail if you would like a copy of this article.
%d bloggers like this: