Share with your CFO
Bain consultants Michael Heric and Steve Beam make a pointed case in this autonomous finance discussion that most finance AI deployments are failing not because the technology is immature but because companies bolt AI onto broken processes instead of redesigning the process itself. One consumer products company operating across 180 countries cut its forecasting cycle from two weeks to one hour per week with better accuracy, but the unlock was abandoning bottom-up manual forecasting entirely, not layering a model on top of it. Roughly 20% of finance organizations are using AI in forecasting today.
What this means for your business
The CFO who has already approved three or four AI pilots and seen modest returns is the exact person this conversation is aimed at. Heric’s admission that early copilot deployments produced poor ROI and mostly generated a bill is unusually candid for a Bain partner selling transformation engagements, which makes it worth taking seriously. The dividing line between finance functions that capture real value and those that accumulate interesting experiments is whether leadership is willing to kill the old process, not just run AI alongside it.
Beam’s rule-of-thirds framing is the sharpest structural claim here: roughly a third of current finance work should simply stop, a third is ripe for AI-assisted process improvement, and a third benefits from speed. That framing matters because it reorders the conversation. Most AI business cases are built around the third bucket alone, which is why the ROI disappointments are so consistent. If your finance transformation isn’t explicitly identifying work to eliminate and work to redesign, you’re spending on speed while the drag stays in place. The counterexample Beam offers, a client with 15,000 AI-generated models whose planning cycles are as long as ever, is the cautionary version of the same point.
The talent signal buried at the end deserves more attention than it got in the conversation. Anthropic usage data shows widespread adoption among accountants even at companies that haven’t sanctioned the tools, meaning your finance team is already using consumer AI on work that touches your numbers. That’s not an endorsement of shadow IT; it’s a leading indicator that the cultural resistance to AI in finance is lower than CFOs typically assume, and that the real friction is process rigidity at the top, not skepticism at the analyst level. Finance leaders who keep framing this as an adoption problem are solving for the wrong constraint.
Concept deep-dive: Agentic AI
Agentic AI refers to systems that don’t just answer questions but take sequences of actions autonomously, like an automated employee that retrieves data, runs calculations, and updates a forecast without a human directing each step. Unlike a chatbot or a copilot that waits for prompts, an agent executes a workflow end-to-end. In finance, that means the difference between a tool that helps an analyst build a model and one that runs the monthly close process on its own, which is why guardrails and defined scope matter so much before deployment.
Based on reporting from Autonomous Finance and the CFO’s Next Frontier, originally published 2025-12-03 03:00:00.
