A survey conducted by the Reserve Bank of India (RBI) found that only a fifth of the financial entities in the country were active in AI adoption.

The survey is part of RBI's FREE-AI Committee Report in which a committee has recommended a framework for responsible and ethical enablement of AI for the financial sector.

The Department of Supervision (DoS) survey found that only 20.80% (127) of 612 surveyed entities were either using or developing AI. 

The surveyed entities included various types of banks, non-banking finance companies (NBFCs), Asset Reconstruction Companies (ARCs) and All India Financial Institutions (AIFIs), representing close to 90% of the asset size. 

The low number of AI adoption among the entities was due to non-adoption in a majority of smaller Urban Co-operative Banks (UCBs) and NBFCs.

In  case of UCBs, no AI usage was reported by tier 1 UCBs, while adoption among tier 2 and tier 3 UCBs remained below 10%.

Besides, only 27% of the 171 surveyed NBFCs have been using AI in some manner. 

Among Asset Reconstruction Companies (ARCs) the adoption of AI was zero, it said.

Notably, while larger public and private sector banks have greater adoption, it was mainly in the form of simpler rule-based models or early-stage exploration of advanced models, it explained.

The findings of the DoS survey was corroborated by another Fintech Department (FTD) survey for this same report. The FTD survey also showed that AI adoption remained low and limited to larger institutions with simpler models that require lower investment and infrastructure, it added.

The report also said that in most cases the use of AI was limited to simple applications such as predictive

analysis, lead generation and chatbots for customer queries.

The report further said that the surveyed financial institutions mainly relied on simple rule based non learning AI models and moderately complex ML models. Advanced AI models had limited adoption.

The most common AI applications used were in customer support (15.60%), sales and marketing (11.80%), credit underwriting, (13.70%), and cybersecurity (10.60%).

"These functions typically involved lower risks, structured flows, predictive outcomes and easier implementation, making them conducive to early AI implementation," it stated.

It also noted that the cybersecurity applications used mostly included third-party enterprise solutions that were easier to integrate with existing systems.

The bright spot was that from the FTD survey, it was observed that out of the 76 entities, 67% were exploring at least one Generative AI use case. 

However, it was observed that most use cases were in an experimental phase and limited in scope, it added.

The RBI report has recommended the formation of a permanent multi-stakeholder AI Standing Committee under the central bank to continuously advise it on emerging opportunities and risks, monitor the evolution of AI technology, and assess the ongoing relevance of current regulatory frameworks. 

It added that the Committee may be constituted for an initial period of five years, with a built-in review mechanism and a sunset clause. 

The report also suggested that a dedicated institution should be established for the financial sector, operating under a hub-and-spoke model to the national-level AI Safety Institute, for continuous monitoring and sectoral coordination.

Further the report said that the RBI may consider allocating a fund for setting up of data, compute infrastructure to support the needs of the financial services sector.

It also recommended that indigenous AI models like Large Language Models (LLMs), Small Language Models (SLMs) or non LLMs designed specifically for the financial sector should be developed.

The report also suggested establishing AI innovation sandbox for the financial sector to enable regulated entities, fintechs, and other innovators to develop AI-driven solutions.

Regulators should also encourage AI-driven innovation that accelerates financial inclusion of underserved and unserved sections of society and other such affirmative actions by lowering compliance expectations as far as is possible, without compromising basic safeguards.

The RBI report consists of 26 recommendations under six categories including infrastructure, capacity, policy, governance, protection and assurance.