5 things ChatGPT (and GPT-4) can’t solve in financial services…yet

There is a ton of buzz about ChatGPT and the know-how’s potential for purposes in customer support, writing and analysis — particularly given the current launch of GPT-4. It jogs my memory of the sooner days of synthetic intelligence (AI) and the thrill round its potential. In lots of methods, the thrill was well-earned and the predictions in regards to the methods firms might apply AI had been spot-on. Machine studying (ML) and AI are serving to to ship extra tailor-made suggestions for ecommerce, supplementing customer support groups with chatbots in locations like LinkedIn, and serving to us all keep slightly safer on the highway with lane steering and emergency braking.

But as with lots of improvements, among the hype went far past actuality. Robots aren’t taking on in the classroom, and to my dismay are nonetheless not capable of tackle all of our extra handbook, tedious duties at dwelling or in the workplace. AI has not ruined the classroom or changed the necessity for individuals to construct product methods, design instruments and supply a human layer on prime of these chatbots when extra advanced points come up.

So after I began studying the hype round ChatGPT, and now GPT-4, I used to be intrigued however skeptical.

After getting an opportunity to play with it, I’ll admit it’s spectacular. Just final week, our CFO was enjoying round with it to assist present some context about our financial projections, and it was fairly spot-on. There is a ton of potential in relation to purposes of this know-how that I’m excited to see materialize.

That stated, we’re nonetheless a great distance from handing things over to AI whereas all of us go sit on a seashore someplace.

When it involves financial companies, there are nonetheless lots of things neither ChatGPT nor GPT-4 can solve, not less than not but. This is as a result of financial merchandise include quite a lot of danger. Financial establishments (FIs) are accountable not only for making certain the security of their prospects’ belongings, but in addition for assembly authorized obligations round know-your-customer (KYC) and anti-money laundering (AML) necessities. FIs even have a vested curiosity in minimizing danger and, consequently, fraud as a result of any misplaced funds can be subtracted from their backside line.  ChatGPT/GPT-4 aren’t but ready to satisfy these crucial danger priorities. Here’s why.  

1. Compliance checks 

Compliance is a crucial a part of each financial companies enterprise. As it needs to be, on condition that firms are dealing with cash for shoppers and companies. AI might help in relation to monitoring suspicious exercise. However, to make sure compliance with confidence, firms additionally want specialists to judge evolving guidelines, decide methods and oversee the compliance program to make sure firms are assembly these necessities. 

2. Making credit score underwriting choices 

Data evaluation has lengthy been part of the credit score underwriting course of, however figuring out the precise insurance policies to make use of to tell what knowledge goes into these choices requires human perception. FIs want to judge their danger priorities to find out what credit score thresholds are appropriate for his or her enterprise. Then, they’ll use credit score bureau knowledge to judge if a buyer meets their credit score insurance policies.

3. Providing a seamless person expertise 

When opening an account, prospects anticipate a seamless expertise that may be accomplished in 10 minutes or much less. To facilitate a frictionless course of with out rising their danger, FIs have relied on things like phone-based identification verification and doc verification, which might mechanically confirm a buyer’s identification primarily based on data they’ve entered throughout the onboarding course of.

However, when addressing points post-account opening, prospects anticipate a extra immersive expertise. Though many FIs use chatbots to assist prospects tackle primary inquiries, if a buyer suspects they could have been the sufferer of a social engineering rip-off, they anticipate to work together with a financial institution consultant on to report the issue.

4. Designing new financial merchandise

Developing new financial merchandise requires a deep understanding of market tendencies, buyer wants and the regulatory atmosphere. It additionally includes making strategic choices that transcend what knowledge alone can inform us. While ChatGPT/GPT-4 can present insights and solutions primarily based on knowledge evaluation, it can not exchange the creativity and instinct of a human designer.

5. Handling a disaster like a fraud assault

While ChatGPT/GPT-4 might help with buyer interactions, fast questions, instructions to assist supplies, and paperwork when an organization is experiencing one thing like a high-velocity fraud assault, they need direct human experience to information them by means of the method. 

The identical goes for stopping fraud assaults. Fraud fashions are useful instruments, however to actually transfer on the tempo of fraud, firms want AI/ML groups to assist guarantee their insurance policies are up-to-date, they’ve the precise datasets in place, and they’re able to take a look at and make updates to their workflows to deal with assaults once they come up.

The way forward for ChatGPT and GPT-4

ChatGPT, GPT-4 and any future updates can be highly effective instruments that may assist financial companies firms in some ways. However, these merchandise aren’t capable of exchange among the higher-touch, extra nuanced components of working a financial companies enterprise.

That stated, firms which are capable of strike the precise stability between automation and human contact can be greatest positioned to realize long-term success by rapidly and persistently delivering worth to their prospects.

Charles Hearn is a cofounder and the CTO at Alloy.