Spreadsheets were the glue between every tool in your stack for 20+ years. Flexible enough to fit anywhere. Now AI agents and workflows are replacing that messy middle layer, and it changes everything.
LLMs are excellent at sounding right when they're wrong.
For finance teams, that's disqualifying. You can't defend board numbers with "the AI said so."
A better model won’t fix that. You need a structure underneath the model that makes the reasoning inspectable.
AI tools are like interns with no context about your business.
"Analyze the P&L" gets a generic answer.
"Go line by line, weekly variance, 13-week cash forecast, COGS by geography, tie it to pipeline" gets you something useful.
AI widens the gap b/w strong & weak operators.
"CFOs should vibe code everything themselves."
"AI can't be trusted near the numbers."
Both positions are useless in practice.
Finance leaders need to go task by task.
Where does AI earn trust here?
Where doesn't it?
That map is different for every business.
"I'm not worried about speed. I'm worried about trust."
Finance teams spend hours validating that QuickBooks matches Salesforce, HubSpot matches forecasts, payroll matches budget.
Speed without validation just gives you faster wrong answers.
Why do forecasts break? Most models assume stability. But real decisions happen during shocks: churn spikes, supply delays, hiring freezes.
Finance teams should help companies rehearse for uncertainty. Make scenario thinking the default, not a nice-to-have. Always ask "what if."
AI in finance works like a new intern. Tell it "analyze the P&L" and you'll get something vague. Give it precise instructions and it delivers. The sharper your input, the sharper the output. Your experience becomes the multiplier.
79% of FP&A teams use AI, but most just automate Excel & polish reports.
https://t.co/xh136J1Lc7: "Few teams are using AI to drive scenario modeling, influence planning cycles or guide cross-functional decisions."
Huge gap between "we added AI" and "AI transformed our work."
Finance teams rebuild the same models over and over. New hire? Rebuild. New market? Rebuild. New board question? Rebuild. The problem isn't Excel... it's lack of memory. People remember why decisions were made. Models don't. We don't need more spreadsheets. We need more memory.
Most finance tools tell you what went up or down. That's table stakes. The real question: "Why did this change?" Finance teams waste hours re-explaining the same variance.
Decisions don't fail from lack of data, they fail when the why behind the data gets lost.
You don't need a 20-person finance team anymore. 2-3 people with the right technology can do what used to require an entire department. Spot trends faster. Catch anomalies earlier. Move from reactive to proactive.
Most AI discourse in finance is people arguing from the extremes.
John Fong at Booth is doing the harder thing: experimenting with tools like Claude & Cowork, and honestly documenting what he finds.
Go follow his work: https://t.co/LDnkWSqrWA
The biggest barrier to AI in finance isn't the tech. It's trust. If you feel like you need to double check every output, the tool hasn't done its job. Transparency and auditability can't be afterthoughts.