Drew Fallon: Harnessing AI to make better financial tools
2019 Finance alum says students should be curious, seek to innovate in specific areas
Drew Fallon: Harnessing AI to make better financial tools
2019 Finance alum says students should be curious, seek to innovate in specific areas
When Drew Fallon graduated from the Farmer School of Business with a Finance degree in 2019, you could probably say he left Oxford already at full speed. He was the first chief financial officer for Mad Rabbit, a tattoo aftercare company started by a pair of Miami University students while they were still in college.
Fallon helped guide the company’s financial growth in those early years. But he said he also found himself frustrated, not with the job or the people, but the technology that was available to help him do his job.
“A lot of the work that I was doing, I just felt like it should be done by computers. The thing about finance is its very rules-based and logical. There's not a ton of interpretation in finance. Its very math driven. There’s a lot of different tools that try to automate a lot of different things,” he said. “But when ChatGPT launched at the end of 2022, I immediately thought to myself, ‘I can't miss this, because this is what I've been waiting for.’”
Fallon said he saw that artificial intelligence could take on much of the heavy lifting for his job. “This thing is going to do accounting, it's going to do finance, it's going to do reporting, it's going to do analysis, it's going to do everything,” he said. I was ready to start my own thing and just went in headfirst.”
He started Iris Finance in the summer of 2023 to be an AI platform for financial modeling and analytics. “We leverage proprietary data orchestration infrastructure. We have our own proprietary data pipelines. We source all preprocessed data direct from Walmart, Amazon, Shopify, Facebook ads, etc, and we process that data in a way that makes it very friendly to be used by AI agents for the purposes of financial modeling and analysis,” Fallon said.
Fallon noted that AI can’t just take a lot of information and do something effective with it on its own. “Giving an AI a massive dump of transaction data and saying, ‘Hey, build me a three-statement model,’ doesn't work. You have to be very selective and opinionated about how that data should be structured, and then the agents can do less data modeling and more data analysis. So, we're sort of a combination between the data layer and the application layer,” he said.
Part of the problem is that AI is powerful, but until recently, it hasn’t been very good at remembering things.
“For example, financial modeling, Finance 303 at Miami. I spent an entire semester learning that workflow, and I probably wasn't even that good at it by the end. There's so many steps and it's a very lengthy workflow. AI has never really had great memory. If you give it something that is too long to execute, somewhere along the way, it'll forget what it's doing, and it'll hallucinate,” Fallon said. “The recent advancements in AI, around Openclaw, Opus 4.6, Anthropic, the memory has just gotten so much better.”
“There really was a big, big change in AI in the last four or six months. If you asked four months ago, ‘How much longer is it going to take you to get this thing to work?’ I would have said probably two years,” he said. “And now, it just works.”
For students looking to become proficient in AI use, Fallon said he’s found that sometimes the best way to learn about an AI or get it to do what you want, is to ask another AI.
“If you're saying, ‘Hey, I want to do an analysis,’ you might find that humans are not actually very good at talking to computers, but computers are very good at talking to computers,” he said. “So, if you explain to ChatGPT, ‘Hey, I'm trying to get Claude to do this thing,’ It'll write you out like a prompt that's computer friendly, very binary and specific. Then you can just paste that into Claude and your Excel workbook to get it to do that same analysis.”
His advice for those who want to be at the forefront of AI? First, be curious.
“Being very, very curious and motivated to understand the latest tech news, because it's a full-time job just to keep up with. I'm sure there's five things that came out today that I haven't looked at,” Fallon said. “You need somebody to be literally trying stuff, because at the end of the day, all this is, ‘How do I get this model to do this thing?’ And it's much easier said than done.”
Second? Seek to innovate in a very specific area.
” I think you have to pick something that the AI is not necessarily that good at, and then you have to figure out how can you engineer it to not just better, but better than the best human,” Fallon said. “With Iris, we don't innovate at the model layer, we innovate at the data layer, at the front end, at the agent layer. The better the models get, the better Iris is going to be, because we're just them plus the proprietary data and a sort of workflow training.”
That innovation and hard work has paid off. Fallon was named one of Forbes’ 30 Under 30 in Food and Drink earlier this year. The magazine said Iris is poised to bring in $3 million in revenue this year and has raised more than $8 million.
As for the future, Fallon said he sees opportunities to use Iris’ work in other areas of business. “We're building this intelligence that right now is being applied to internal finance operations at consumer brands. But we have a couple of private equity funds that are now using Iris because the agents are running the diligence process,” he said. “It's the same intelligence that can be applied to so many different industries. We are building the financial intelligence layer for anything consumer related.”