CFO Daily · Friday, September 25, 2026

Hi, this is Koko from Koko Knows.

Let's get right into it, because the sharpest signal today is a contradiction, and it's one every finance leader needs to sit with. Bank of America just announced it's doubling its AI budget for next year. Same week, Accenture research finds that only twenty percent of bank leaders are seeing widespread, sustained value from their AI initiatives so far. Read that again: spend is accelerating while proof is not. That gap is no longer a side conversation for your innovation team, it's now the central budget question on your desk.

Zoom out and you'll see this isn't just a BofA story. Gartner is projecting worldwide AI spend will hit two point seven trillion dollars in 2026, up nearly fifty percent year over year. But here's the part that should change how you read every vendor renewal crossing your desk this quarter: Gartner's own analysts say a lot of that growth is relabeling, not new money. IT budgets getting an "AI" sticker slapped on them at renewal time. So before you sign off on next year's AI line item, have FP&A go vendor by vendor and force the question: is this genuinely incremental spend, or is this an infrastructure refresh wearing a new name tag? You want a real number, not a rebrand.

Third thing worth your attention today, and it's a talent story hiding as an HR story. Job postings asking for AI skills are up roughly eight times since 2022. Compensation for those roles hasn't moved anywhere near that fast. That's a wage-arbitrage window, and it's closing. If you've got controllers or FP&A analysts who are AI-literate, or you're trying to hire them, there's a window right now to lock in talent before pay catches up to demand. That's a build-versus-buy decision worth having this week, not next quarter.

Now, here's where I want to connect the dots for you, because these three stories are actually one story. Capital is flowing into AI faster than the return is showing up, budgets are getting relabeled to make growth look real, and the people who can actually make AI work are underpriced relative to demand. If you're a CFO, that combination should make you tighten, not loosen, your approval criteria. Tie any AI budget increase to specific, dated ROI checkpoints. Not "improved efficiency." Not "enhanced capability." A number, a date, an owner.

One more thing I want on your radar, because it's the kind of risk that doesn't show up until it's too late. There's a vendor lock-in framework circulating right now that everyone signing agentic AI contracts should read before their next signature, not after. It points back to Builder.ai's collapse in 2024, where enterprise clients discovered they couldn't get access to the governance documentation or even the source code behind production systems they were running on Builder's infrastructure. When that vendor went under, those clients were stuck. If you're running any agentic workflow, especially in finance operations, through a vendor's infrastructure rather than your own, ask today: what happens to this workflow if this vendor disappears next year? If the answer is "we don't know," you need a data and code portability clause in that contract before you renew, not after something breaks.

And a smaller flag worth filing away: a Blue Cross Blue Shield study found AI-assisted documentation added nearly a billion dollars in costs, largely by generating new billable codes, not by patients actually being sicker. That's a caution for any of you using AI to auto-generate expense codes or documentation in your own shop. Volume growth from automation can look like productivity when it's actually just the system finding more things to bill or code. Worth an audit if you've rolled anything like that out.

On the cost side, there's actually a bit of good news moving underneath all this. Anthropic and OpenAI both cut model prices by roughly forty to fifty percent on the same day. That should be pulling your AI unit costs down right now, if procurement is renegotiating contracts to reflect it. But don't stop at price per token. The bigger leak for a lot of enterprises isn't the sticker price, it's idle GPU capacity sitting behind bad orchestration. Before you sign any new capacity commitment at today's lower prices, get an audit of what you're actually using versus what you're allocated. That's real money hiding in plain sight.

So here's the bottom line for today. Demand the vendor-by-vendor accounting on AI spend before you approve next year's budget. Get portability clauses into any agentic contract touching production workflows. And move now on AI-literate talent while the pay gap is still in your favor. Three moves, all doable this week.

That's the rundown. I'm Koko, and this has been Koko Knows. Talk to you tomorrow.