Generative AI for financial services is a "valuable tool, but yet to be fully disruptive"
Let's look at what's holding back the GenAI revolution
GenAI scores 100% in Cost Reduction but also nearly 100% in competition! For most financial services GenAI will be revenue positive, except in Insurance where increased competition will be brutal.
UBS and MIT look at Generative AI (GenAI) in financial services (finserv) and find that the transition to GenAI-enabled banks won’t be as fast as some predict!
I’ve covered the risks banks face with GenAI before, see this article:
So in this article we will look at the challenges banks need to overcome to deploy GenAI!
👉TAKEAWAYS: Challenges to deploying GenAI
Data quality: GenAI requires high-quality data to train models and make accurate predictions. Data bias is a real issue!
Regulatory compliance: Financial service firms operate in a highly regulated environment and must deal with new regulations and ethical concerns.
Ethical considerations: GenAI has the potential to perpetuate biases and discrimination, and financial service firms must ensure that its use is ethical and transparent.
Talent shortage: There is a shortage of talent with expertise in GenAI. Good luck hiring a team!
Infrastructure: Financial service firms must invest in the infrastructure needed to support GenAI.
Cost: Deploying GenAI is expensive and this is before regulatory and compliance costs are factored in.
Resistance to change: GenAI represents a significant change for financial service firms, and there may be resistance to adopting new technologies and ways of working.
EXPLAINABILITY!!! The black-box nature of genAI can make it difficult to identify and correct errors or biases in the system. Imagine explaining GenAI's decisions to regulators and customers.
HALLUCINATIONS!! For now, there is still no solution for this and it’s a big problem!
Competition risk is most likely the biggest benefit for consumers. In a GenAI-enabledh financial world pricing will likely be far more transparent between offerings. This is true even if GenAI makes understanding how prices are derived more opaque.
👊STRAIGHT TALK👊
The craze surrounding GenAI is a problem. It is reminiscent of the craze surrounding blockchain in 2016-17 that predicted that a clunky ledger would do everything, even cure cancer.
Yes in 2017 people actually wrote stuff like this. This is what hype can do!
It’s not that GenAI or blockchain are bad; they are fantastic technologies. However, both need years of development. This year, we saw fabulous bank uses of blockchain seven years after the peak of the craze. I bet that GenAI will need five years before it hits banking with its full potential.
The two big problems for GenAI are that it is unexplainable and that it has hallucinations. GenAI is famous for hallucinating answers, but I believe this is a technical problem that can be solved.
This leaves explainability that can’t be solved and is built into every AI and GenAI system. The problem is that regulators and clients need answers to the question “why,” and GenAI’s lack of explainability is a real problem.
Let’s make this real. A credit rating agency in the US was sued by minorities who noticed that their credit rating dropped when moving to ethnic neighborhoods. The credit rating agencies admitted that their algorithms were based on postal code, a prohibited practice called “redlining” and paid out. The point is that the courts and clients needed a why, and they got an answer because algorithms are explainable.
When an AI “red lines” your loan or other application, how will you get an answer? This isn’t theoretical; it's a real problem for clients and regulators.
I love GenAI, but for it to be “fully disruptive,” it will take at least five years.
Stop all the GenAI hype and deal with it!
