For most of his career, Tom Mancini thought finance was about numbers.
Tom started his career in investment banking before moving into private equity. Like many finance professionals, he spent years immersed in analysis, valuation, and deals. Then one day, he found himself at a consumer and luxury startup accelerator with a newfound perspective. For the first time, working alongside founders and operators, Tom saw finance leave the spreadsheet behind.
"Finance stopped being spreadsheets for me," he says. "I saw financial thinking applied directly to real products and day-to-day decisions."
Watching teams wrestle with messy business problems taught him that the real value of finance isn't building perfect models. It's bringing clarity to ambiguity. That tension—taking something complex and making it intuitive—is what continues to motivate him today.
Teaching AI to reason, not calculate
Most people assume finance expertise in AI means checking calculations or verifying facts.
Tom's work is much closer to mentoring a junior investment analyst.
"It isn't about whether a model is simply right or wrong," he explains. "It's about whether the reasoning actually holds."
That means looking beyond polished outputs to examine the assumptions underneath. Is the conclusion more confident than the evidence supports? Does the analysis connect cash flow to returns in a way that reflects how businesses actually operate? Has the model overlooked an important business reality?
"The hard part isn't spotting the obvious mistake," he says. "It's spotting the answer that looks right enough to be dangerous."
Those are the kinds of judgment calls that rarely show up in benchmarks but matter enormously in practice. The goal isn't simply to produce correct answers. It's to teach AI to reason the way experienced investors do.
The underestimated skill behind frontier AI
Tom expected his raw finance background to translate well to AI work; what he didn’t expect was how important collaboration would be.
"I underestimated how central collaboration would be," he says. "When your team is spread across the world and the project is moving quickly, cohesion becomes a skill in itself."
Years of working across distributed teams prepared him for that environment. Aligning people quickly, taking ownership, and keeping projects moving have become just as valuable as financial expertise.
Seeing AI from the inside
Working directly with frontier models has made Tom both more optimistic and more discerning about AI.
From the outside, it's easy to think of AI as either a threat or a shortcut. After working closely with these systems, he sees something different.
"It's a new way of working that quietly forces you to get sharper in your own field," he says.
In finance, today's models are remarkably capable. But capability isn't the same as judgment. The strongest results still come from experts who know when to trust an output, when to challenge it, and when business context changes the answer entirely.
Rather than replacing expertise, AI has raised the standard for it.
"It feels like standing at the start of a new professional standard that's still being written," Tom says. "I'd much rather help shape it than watch it take shape from the sidelines."
What AI still misses
For all of AI's strengths, Tom believes some of the hardest problems remain surprisingly human.
One example is irony.
"It depends on context, timing, tone, subjectivity, and a certain elegance in saying something without quite saying it," he says.
Building the future from the inside
"Working on Mercor projects is one of the most fascinating ways to understand how these models actually think—from the inside rather than through headlines."
It's easy to imagine frontier AI as a story about algorithms. Tom remembers it differently.
He talks about the people he's worked with across projects, the relationships that formed despite the distance, and the conversations that continued long after a particular engagement ended.
"It's unusually enriching."
The models may be getting smarter but the work? That is still fundamentally human.