AI's Effect on Compliance

Recently read an interesting article regarding the future for those who work in “back office” roles”, especially in compliance. According to this article, big banks, such as HSBC, Deutsche Bank, and JPMorgan, spend well over $1bn a year each on regulatory compliance and controls.Thus, banks have turned to artificial intelligence to deal with compliance issues.

According to the article:


Senior executives from financial companies, 49% said they expect their firms to use AI for risk assessment within the next three years. Another 29% anticipate that their firms will apply AI to learn more about their clients and to prevent money laundering. Another 26% expect that AI will be used to help with regulatory as well as risk and compliance issues.

Will AI replace compliance in the future? How about other back office roles?

Personally, I do not think this is the case, I only think that AI will only be used as a tool for compliance departments to make work easier, but I am interested in seeing how AI will affect the overall banking industry.

 
Best Response

As someone who works in AI, this is partly true. AI's can detect a lot more nuance/context than you think. NLP has come a long way. You just need to be using the right tools.

To understand the difficulty with using AI for something like compliance, you have to look at type 1 vs type 2 error (also known as false positive vs false negative). The most common difficulty that AI's face in these types of situations is too many false positives. AI's are quite good at detecting fraudulent behavior when there actually is fraudulent behavior. Unfortunately, AI's will often flag non-fraudulent behavior as fraudulent (generally speaking). This leads to over-auditing. This is partially mathematical and partially by design. False negatives, in most instances, are more catastrophic than false positives. Sure you waste more time if you get a lot of false positives but at least you aren't missing frauds.

At this stage an AI could not fully take over a compliance department. However, using AI you can greatly reduce the number of cases that compliance would have to look at. That's where the real value of AI comes in. You will almost never get false negatives but the cost of that is you'll end up having to sort through some more false positives in order to identify where the real problems are.

 

A valid point Deep, I suppose I'm just considering AI in the RIA compliance world--i.e. does this graphic constitute investment advice, does this phrase constitute a guarantee of investment returns, is this phrase potentially misleading, etc. You're probably right about applying it to different areas however. That said, I still hate dealing with the Business Prevention Unit (compliance) on any level.

 

It really depends on what people mean by AI. Most banks are using some sort of expert system, a bunch of hard-coded rules arranged as nested if-then lists, and that's a crude tool that will surface a lot of false positives. Which leads to a huge waste of time and money on the part of the bank and its analytics team.

The best thing to do is combine a rules engine with another filter based on some sort of unsupervised learning, like a neural network. With the manually coded rules, certain unusual behavior which banks know is legit can be allowed to proceed unflagged, other unusual behavior that is known to be illegal can be bucketed as illicit. The real problem is that hard-coded rules provide bad actors with an excellent feedback loop. You test, get blocked, and test again until the transaction goes through, at which point you exploit the breach.

One advantage of deep neural networks is that they can be trained in real time to detect new types of anomalous behavior, the unknown unknowns. You don't have to wait for a domain expert to come up with a new rule to catch fraud or money laundering. The data itself teaches the nets what's normal and what's not.

 

AI is a great tool for analysis but it won't be replacing people in compliance type roles any time soon. And you're right about false positives--look at what happened with Soundcloud. They use Google's Content ID for copyright detection, and a lot of time end up tagging songs that are similar but are not actually copyrighted, or, in one case, it went as far as deactivating an artist's account because it detected the songs were copyrighted even though the artist who owned the copyright was the one uploading them on his official account. Problem apparently was that the account hadn't been cleared through the label side of things. So yes, AI is great for analysis but, at least for now, terrible for execution in context-driven sonarios.

 

Picking out issues from a dataset sure. However, relationships with client will always require some degree of intuition. Somebody in compliance will know that sometimes you need to bend the rules to maintain the relationship.

I should emphasize internal regs not legal...

Only two sources I trust, Glenn Beck and singing woodland creatures.
 

Optio voluptatibus expedita animi officiis. Magni adipisci sit omnis laudantium vel tempora repellat. Est dolore eligendi sint ut occaecati harum. Illum et ut sunt quos.

Qui optio qui minima provident. Sapiente sint commodi veniam quisquam veniam. Et sint accusantium sapiente est molestias hic ipsum.

Itaque dolores repudiandae et explicabo consequatur aut tenetur consectetur. Nam sint labore ratione et sint repudiandae aut. Velit blanditiis voluptas nisi eum tenetur necessitatibus. In quis dolores iste. Debitis architecto autem nam numquam nostrum dolor qui.

Cum quasi quis distinctio repudiandae adipisci. Sed magnam perspiciatis libero consequatur a dolorem quia. Illum pariatur iusto voluptates quaerat accusantium. Vitae soluta at a quisquam in. Reiciendis officia ad error modi. Et nemo iure nam magni. Id similique nesciunt non dignissimos enim esse et.

Career Advancement Opportunities

March 2024 Investment Banking

  • Jefferies & Company 02 99.4%
  • Goldman Sachs 19 98.8%
  • Harris Williams & Co. (++) 98.3%
  • Lazard Freres 02 97.7%
  • JPMorgan Chase 03 97.1%

Overall Employee Satisfaction

March 2024 Investment Banking

  • Harris Williams & Co. 18 99.4%
  • JPMorgan Chase 10 98.8%
  • Lazard Freres 05 98.3%
  • Morgan Stanley 07 97.7%
  • William Blair 03 97.1%

Professional Growth Opportunities

March 2024 Investment Banking

  • Lazard Freres 01 99.4%
  • Jefferies & Company 02 98.8%
  • Goldman Sachs 17 98.3%
  • Moelis & Company 07 97.7%
  • JPMorgan Chase 05 97.1%

Total Avg Compensation

March 2024 Investment Banking

  • Director/MD (5) $648
  • Vice President (19) $385
  • Associates (85) $262
  • 3rd+ Year Analyst (13) $181
  • Intern/Summer Associate (33) $170
  • 2nd Year Analyst (65) $168
  • 1st Year Analyst (198) $159
  • Intern/Summer Analyst (144) $101
notes
16 IB Interviews Notes

“... there’s no excuse to not take advantage of the resources out there available to you. Best value for your $ are the...”

Leaderboard

1
redever's picture
redever
99.2
2
Betsy Massar's picture
Betsy Massar
99.0
3
BankonBanking's picture
BankonBanking
99.0
4
dosk17's picture
dosk17
98.9
5
GameTheory's picture
GameTheory
98.9
6
Secyh62's picture
Secyh62
98.9
7
CompBanker's picture
CompBanker
98.9
8
kanon's picture
kanon
98.9
9
pudding's picture
pudding
98.8
10
bolo up's picture
bolo up
98.8
success
From 10 rejections to 1 dream investment banking internship

“... I believe it was the single biggest reason why I ended up with an offer...”