a16z Fintech & Citigroup Talk AI, & Spending on AI Estimated To Skyrocket
Here's what's going on in AI x Banking this week.
Hi everyone! Hope this week’s going great. We’re off to the Datos Financial Crime conference next week where Greenlite CEO Will Lawrence is speaking about automation in compliance. If you’re in Charlotte or will be at the conference, please let us know!
Lots to get through today, so let’s dive right in:
New Research Shows Big Bank Spend on AI
A new IDC report says that spending on Gen AI could reach $202 billion by 2028, with 20% of that (or approximately $40 billion) being led by financial services firms.
Before we jump in, it's important to note that there are going to be a lot of estimations and research about AI spending over the foreseeable future. Research firms put out research like this all the time—IDC is a reputable firm so their research is worth mentioning, but always take such reports with a few grains of salt.
But, there are a lot of facts that back up these assertions.
A KPMG report surveying 250 executives at companies making more than $1 billion a year also came out last week too. 83% estimated to increase their AI budgets over the next 3 years.
It’s also true that AI and Gen AI projects are moving from tests to pilots and full implementation. The KPMG report also says that 61% of respondents plan to expand the scope of current Gen AI projects, while 51% want to integrate Gen AI into new business functions.
All signs are trending up for AI’s future inside banking services. More money, energy, and time is being devoted to these projects, and with it, likelihood that banking will be fundamentally transformed by Gen AI.
a16z Fintech Interviews Citigroup’s Head of Strategy
We wanted to highlight a great new interview between a16z General Partner David Haber, partner Marc Andrusko, and Citibank’s Tim Karpoff—it was filled with interesting tidbits around the bank’s vision around integrating tech and Gen AI.
Below we’ve collected some highlights about AI, but highly recommend reading the transcript or listening to the podcast too:
AI’s Impact on Banking Functions: Everyone’s been hypothesizing around use cases for AI in banking. Citigroup’s Karpoff, who’s seeing these changes firsthand, had some thoughts:
Payments can be much more efficient and be more automated. Citi is already using AI for fraud scoring. "Operational losses from fraud will come down further and further as a result of the deployment of AI, machine learning,” Karpoff says.
Capital formation, or the process of raising capital, can benefit from AI—helping teams identify new origination opportunities, use payment data to better understand clients, improve the timing of client interactions, and enable more sophisticated pricing. Just being able to free up people to spend time on relationship building versus crunching number can have a huge impact.
App development will see a “step change, and that really will be transformative,” according to Karpoff. It’s already the case outside of banking, with people being able to put together whole apps through Anthropic’s Claude chatbot.
Areas of Slow AI Adoption: This is a great underdiscussed topic—where will AI be slow on the uptake?
Mortgage Origination: While folks think lending can be improved through Gen AI, that might not be the case. Non-US banks have already rolled out AI-enabled solutions but there’s been limited customer uptake so far. "For the largest, thorniest transactions, people still want a human being." Getting a mortgage is a significant life event that people will still want to guided through.
Being Cautiously Optimistic:
Optimizing operations is great and all, but at the end of the day, public shareholders really care about generating revenue. That path is still uncharted. "Longer term AI has got to be deployed in a manner that's going to generate that [growth],” Karpoff says.
It’s also important to move fast but not too fast, to understand the impact gen AI can have on consumers and the company as a whole. On top of that, the regulatory landscape is still pretty murky. That’ll have huge implications on the rollouts of AI initiatives.
Thinking About Smaller Models
We really enjoyed this interview with Databricks’ Russ Rawlings, Enterprise RVP in the UK, about Gen AI and finance.
One point that Rawlings really wanted to get across was the idea that smaller models are probably going to be better and more utilized by banks.
“When people use the term “LLM”, they usually think of the huge consumer chatbots that have been at the helm of the AI conversation. But financial institutions need a smaller, more secure LLM…A financial enterprise’s model does not need to know anything about, say, celebrities; this data is irrelevant. The model needs to help organization with what matters to them, like determining risk, minimizing fraud, or providing more personalized experiences for customers.”
A really great point—financial focused model’s don’t need to know the plot of Othello. They really just need to be trained really precisely on banking functions. The interview goes on to mention how custom open-source models can be a potential solution—they’re already smaller, so will be easier to run, and high quality datasets will yield better, more accurate, results.
As banks and institutions spend time thinking about how to leverage LLMs, its important to think outside the box and keep track of developments to build the best solution possible.