GenAI Crushes it as a Compliance Officer

A Standard Chartered-backed research project maps the commercial impact of regulatory changes and comes up smelling of roses.

The compliance team. For some, it is a much-hated, surly appendage representing the decline of contemporary wealth management. For others, it is a critical corset keeping the wayward private bank in line, and a client’s assets safe and secure.

Now, after decades of an uneasy truce between the front line, management, and the various lines of defense, there is potentially something new looming in the general risk and control schematic. 

Completed Successfully

To wit, a Standard Chartered media release issued on Wednesday reported it had successfully used AI for industry-specific compliance matters.

According to the bank, its innovation, fintech, and investment arm completed a project with Singapore government-backed institute A*STAR IHPC to automate the complexity of mapping industry rules and regulatory changes to supposedly authorized entities.

Real-World Challenge

In detail, what it did was look at the «real-world challenge» of supervisory change in financial services by optimizing large language models with retrieval-augmented generation that were then able to analyze specific business implications from shifting regulation.

«The project achieved enhanced response quality, improved answer accuracy, and reduced biases,» Standard Chartered indicated.

Bank of Compliance Past

This is a bigger step than most potentially realize. It may leave some speechless, but keeping a log of regulatory changes in each jurisdiction and doing something about it has been the bane of compliance and control functions in financial services for decades now.

It is a thankless job that in the early to mid-2010s tended to be thrown around with large dollops of unnecessary name-dropping and political mudslinging. It usually ended up being a name and shame game with a rotating cast of front-liners, managers, external consultants, legal, compliance officers, and even operations.

Better Responses

Now, the Standard Chartered project comes along and appears to maintain that GenAI is a capable hand in such matters. If it performs as well in a live environment, it is practically guaranteed to improve response quality, the first of the three improvements the financial institution pointed out in its release.

That leaves us with the other two - improved accuracy and reduced biases.

Lack of Knowledge

If the project can also master that feat sustainably, we are on to something spectacular here, as a key weakness of many compliance officers is the lack of detailed business knowledge. 

The more versed they are in the arcane art of being the internal ordinance whisperer, the more it correlates with a hefty sticker price and stints in many different types of banking, mixed in, of course, with spells at external consultancies. 

No Questions Wanted

That invariably means blunt, ham-fisted blanket yes or no's regardless of the exact nature of the portfolio, client, or even the individual action, and potential breach. 

Some of what they do is logical. They must protect their backs and don’t want to face up to regulatory or audit questions afterwards, usually a larger fear than the one posed by what they see as an interchangeable troupe of internal executives and the front-line. A future operational GenAI, hopefully an emotionless and hallucination-free one, would do away with all that.

Less Bias

The last is the reduced bias. It is an achievement that can’t be rated highly enough if it makes the grade in practice. Compliance departments and individual officers tend to develop good and bad relationships with bankers, and this can become dysfunctional over time.

Some front-line staff become internally notorious for testing limits repeatedly while others toe the line, and this knowledge oozes like corporate osmosis from one officer to the other, whether old hand or newbie. 

Unchecked by Management

It can easily become an unwarranted bias if left unchecked by management, which it frequently is, and GenAI here also appears to be the euphemistic banking white knight in shining armor. 

Browbeaten banks and bankers complaining about growing compliance burdens will be able to prompt an objective, always-on tool and get the answers back - without any of that pent-up, passive aggression.