AI

Our Collective Wisdom about AI

Published

January 15, 2024

To get the best of Artificial Intelligence, we need wisdom, and even better than one person’s wisdom is the collective wisdom of many. With this objective, Banking Frontiers did a series of polls on a variety of AI-related topics, and invited experts in the field of BFSI and IT to vote and to share their expert comments. We present the voting analysis in the form of graphs alongside their expert comments. We hope this collective wisdom will be useful to boards and departments and teams and managers as they journey towards the AI frontier. Many thanks to all the participants for contributing to the collective wisdom about AI.

AI – Impact on Customer Engagement

“AI has capabilities to hyper personalize for the customers. The impact will be significant in corporate banking space.”

Aditya Agarwal, Principal, Cedar Management Consulting International

“Intelligent personalized customer engagement using Generative “AI could result into CX which in turn could translate into customer loyalty, up sales, cross sales, and new customer acquisition.”

Divyang Shah, Founder & CEO, Electrosoft

“Generative AI could be used to solve some of the Bottom of Pyramid customers concerns. Multilingual, medium agnostic (written, spoken, emojis) customer engagement plan could be a game changer.”

Charu Dutt Sharma, VP – Business & Sales, Finfactor

“I think the most correct way could be to find or ‘invent’ most ‘human’ way to do it.”

Vishal Powale, Founder, Searchlight Advisors

“Generative AI will help towards not only personalized customer engagement but also will ensure meaningful, contextual and multi-lingual services.”

Pravin Gupta, Founder & CEO, CredenTek Software

“Optimization is the goal for better engagement and cross sell.”

Anurag Srivastava, CEO, U P Coop Banks Federation

AI Impact on Employment

“AI could be seen as a game changer if we approach with the intend to understand the possibilities, redesign future-of-work to leverage this technology as a complementing partner for humans / employees. This would then free up human capacity to focus on high order work, be creative and expand horizons.”

Ambal Saravanan, HR Strategic Consultant, Intellect Design Arena

“Great potential of simplified and more efficient pre-processing of data before presenting them for processing to the mid management. This will surely increase ease, efficiency, accuracy and productivity.”

Divyang Shah, Founder & CEO, Electrosoft

“I think it is already disrupting the industry. My concern is over the uncensored use of data by ChatGPT and lack of compliance/regulations. Therefore, the impact or ChatGPT at present is immense if you consider the content and the communications industry. But this can be a short-term haul as happens with any new tech.”

Sonal Desai, Editor, WriteCanvas

“If AI can make employees productive, the number of customers served can go up.”

Anurag Srivastava, CEO, U P Coop Banks Federation

“Generative AI is being used for assessment of profile for a particular GD.”

M D Agrawal, Program Director, AIMA

AI – Best Role to Maximize Ethics

“CIO/CTO has to practice all IT systems related ethical practice in day today deliverables and operations overall, but other Chief Risk, compliance and Ethics and functional head can second to this.”

Mohan Kumar Venkatesan, Senior Vice President & Head IT Infrastructure & Application Services, Axis Bank

“There has to be a separate Ethics executive to oversee business & non-business processes & practices across the board in any financial institution considering the barrage of information, misinformation & mal-information floating around in the Internet-powered world today. Also, a data privacy law will mean a lot of reworking of data, customer & tech strategy – an overhauling of systems to ensure bias redundancy, diversity + inclusiveness as well as data sanctity.”

Ushamrita Choudhury, Chief Operating Officer, rhizodesic LLP

“Organizations would need a separate vertical called – Ethics Committee of diverse professionals to steer it. There is too much power in the hands of executives leading the day-to-day function.”

Poonam Vijay Thakkar, Mentor, NITI Aayog

AI – Best Tech Combination to drive Innovation & Transformation

“The BFSI industry faces disruptions from the growing impact of Insurtech and fintech, driven by technological advancements. AI, with a substantial market presence, is in high demand due to disruptions caused by Generative Artificial Intelligence (GEN AI), reflecting increased acceptance of these technologies.

The convergence of Augmented reality (AR), Virtual reality (VR), and artificial intelligence (AI) is set to transform the BFSI industry. This fusion of technologies will revolutionize product sales, claims processing, and risk evaluation, enhancing customer service, efficiency, and security.

Current use cases demonstrate AR/VR/AI deployment, with substantial growth potential in the next 3 years. AI-powered AR applications streamline property inspections, refining risk assessments and identity verification. VR-based simulations aid customer understanding of risks, facilitate product purchases, enable personalized experiences, and provide comprehensive employee training.

These technologies promise innovations across fields, supported by IOT devices, 5G infrastructure and faster computing chips.

It is crucial to identify the purpose, right use cases and benefits before venturing into the competitive market taking into consideration the inherent security risks it poses. Well-thought-out use cases, backed by extensive research, are essential for achieving favorable outcomes.”

Tabrej Khan, Sr VP – IT, Anand Rathi Insurance Brokers

Fixing AI Bias

“Bias in an AI system could be due to training Data or the Algorithms or Overfitting (Model Generalization) or Design and deployment bias and finally human bias. We need to understand the source of the bias before we try to fix it. Depending on the function of the AI, some biases are intentional and may not need to be fixed. The best solution to identify and fix AI bias is to build Explainable AI. Data bias is easier to fix, the other biases are progressively tougher and in some instances you are better off starting from scratch.”

Chandramouli Pandya, Chief Technology Officer, Humane Technologies

“There is no one answer which fits all. Starting point would be data, and then look at how algos use / distinguish them and then act.”

Surya Prasad, Co-Founder, Affinsys AI

“A loss function can be designed that penalizes the model for making biased predictions.”

Srimathi Sundara Rajan, Managing Director, SunSmart

“Recognizing Bias in an AI system itself if an experiential learning for the user-groups, and their responses would normally be between ‘Changing the model/algo’ or ‘Use alternative training data’ as the poll suggested. The need to address the ‘different types’ of biases – 6 or N different types of biases, depending on the literature you follow – and has to be driven by objectives. It would be wonderful if all the techniques could be simply bundled into an ‘anti-bias’ solution, but we all know that there is no ‘one size fits all’ which is particularly true for AI/ML. Thus, I would just sum up that no single option works in isolation, but probably is more purposeful when in a ‘bundled fashion’.”

Sridhar Kalyanasundaram, Trainer, Mentor & Consultant in Enterprise Risk Management

Reducing AI Bias

Aligning AI Initiatives

Choosing an AI Partner

“Although companies are made up of people, companies restrict themselves to Object Governance. Whereas AI Techs, I presume, directly are of Subject Governance. Illustratively, a hacker is not the brainchild of companies but of some weird people. Subject Governance is a must for companies to practice but are not doing it. Whereas AI Techs if they bring in SOS Governance [Subject-Object-Self] principles individual’s performance can be gauged. What is critical for AI is not the output but the input, how men and women conduct themselves, not only reporting Fiscal [Object] but their conduct as to adhering to Ethical-cum-co-responsibility [subject-self] factors.”

Jayaraman Iyer, CEO & Founder, IBCM Research

“Horses for courses! Call has to be taken depending on the scale, use case, timeframe and budget. Once you have clarity in terms of these question choices become clearer and easier.”

Piyush Saxena, Sr VP – Cloud Business & Offering Management, HCL Technologies

“Financial organizations are sole source of transaction data of buyers and sellers. India needs to take long and fast strides to leverage each of bigtech, fintech, AI and IT services at different touch points. Convergence of these would sharpen our ecosystem as a whole. Big Tech – in managing huge data repository, creating unique customer profiling to pass advantages to customer for relations with FI. Fintech – the point of transaction execution, needs to be a revenue earner for users and not a cost. AI – to help organizations improving service touch points, knowing pain points before it occurs to customers creating a delight, in spite of pain. Finally, the IT Services – our major market is unorganized and ‘unskilled’ without a cohesive data repository tool. IT services can help solve that effectively and connect them. The challenge is who takes the onus of this convergence into the ecosystem.”

Ajay Prasad, Director, Rise Retail & Payment Solutions

Generative AI: CXO Skills

“The ideal answer would be a combination of Creative, Analytical and Technological skills. A creative and analytical mindset will help find new and innovative ways to use this new tool in their arsenal and the technological skills would help understand the limitations of Generative AI especially in the Fintech world. Concepts such as AI Hallucination which has significant impact in the Fintech world are difficult to understand without technological skills.”

Chandramouli Pandya, CTO, Humane Technologies

“Though I believe more in creative skills, other skill sets are equally important.”

Shiv Singh, Executive Director, Indian Bank

Source of AI Improvements

“AI will produce results (outcome) depending on the models set up with many iterations. Empirical evidence has shown the Business Users who can perfect the models will lead to AI producing consistent and results which are meaning to the User. Once this happens the numbers of Users will exponentially increase. So, in the survey, while leveraging more data is the ticked solution, I would clarify by saying that the human user factor in stitching that data to arrive at the solution model is what is important. As an example, MSME lending digitization model is sub-optimal because the permutation and combination has yet to be perfected. To elucidate, the reason why MSME score card has not been perfected is the absence of good credit profilers who can define the data tables used to search!”

Shantanu Ghosh, Independent Director and Advisor, Multiple Organizations

“In my opinion, AI is still in the development phase despite the significant progress made in the current fourth industrial revolution. It is necessary to involve human experts to ensure the proper evolution of AI, so that it can assist us in performing multiple tasks efficiently. I believe that an important element for the proper development of AI is to adjust or incorporate algorithms to prevent, limit, or warn of the possible negative effects on humans that could be caused by AI’s response or action. It is also important to incorporate the complexity of human reactions, which are diverse and multiple across regions, countries, and even within countries. This is a complex issue because the same groups of people can react differently to the same event at different times and depending on the context. Similarly, at a personal level, our reactions could vary depending on our mood. Therefore, it is advisable to include the delivery of possible consequences scenarios in the development of AI. Other important challenges are data privacy, security, ethical concerns, transparency and explainability in AI development.”

Ronald Omar Pinto Ribera, Independent Consultant, Bolivia

“I think access and affordability is going to be the biggest challenge in times to come. Other things like training humans and generating synthetic data will be taken care to a large extent. Today training a LLM model costs millions. Cost of running prompts is also prohibitive for many organizations. Hyperscalers are not bringing down the cost anytime soon. Another one I think will having a clear, strong and consistent governance process. This can have a huge impact on AI adoption timelines.”

Piyush Saxena, Sr VP – Cloud Business, HCL Technologies

Investment Bias

“AI will start coming up with its own opinions based on the provided data sets. Biases favouring specific industries, regions, or demographics over others to receive financing, can be tough to rectify.”