A few weeks ago, I did something most fractional CFOs probably wouldn’t admit to: I tried to replace part of my own workflow with AI, and it failed.
Not catastrophically. Not in a way that hurt a client. But it failed in exactly the way that matters most in finance: quietly, and in the details.
Here’s what happened, what I learned, and why I think it’s directly relevant to you: the professional services operator who’s been thinking about getting finance off your plate for good.
—
## The Experiment
I set out to build what’s called a “Skill”, an autonomous, reusable AI process, to handle a recurring data cleanup task. The idea was straightforward: document the process once, hand it to the model, and let it run. Less manual work, more capacity for higher-level thinking.
The first output looked clean. But when I reviewed it, something was off. The AI had missed critical formatting details — the kind of thing that wouldn’t just look wrong in a report, it would silently break the downstream process that depends on it.
When I pressed the model on why, the answer was clarifying: I hadn’t actually built an autonomous workflow. I had written a slightly more complicated set of instructions. The AI executed exactly what I told it to do. The problem was I hadn’t told it everything, because some of what I know lives in experience and pattern recognition, not in documentation.
## Why This Matters for Your Business
If you’re running a professional services firm: consulting, marketing, architecture, staffing, whatever the discipline, and you’re still managing your own finances, this story probably sounds familiar in a different form.
You’ve tried a tool. Maybe QuickBooks, maybe a spreadsheet system, maybe an outsourced bookkeeper. It handled the basics fine. Then, a client billed differently than expected. Or a project ran over. Or you had a revenue recognition question that the software couldn’t answer. And suddenly you were back in the weeds, manually fixing something that was supposed to run itself.
That’s not a technology problem. That’s a judgment problem, and judgment is the one thing no tool, AI or otherwise, comes pre-loaded with.
## The Rule That Runs Every Good Finance Operation
In finance, we operate by a principle borrowed from arms control negotiations: **trust, but verify.**
It applies to audits. It applies to month-end reconciliations. It applies to the P&L your bookkeeper sends you every month. And it applies — absolutely — to AI.
What my experiment reminded me is that this rule doesn’t go away when you adopt better tools. It shifts. The question moves from “did someone do this right?” to “does someone know what right looks like?” That’s a more skilled role, not a lesser one. And it’s exactly the role that breaks down when a busy operator is also trying to run their own books.
The professional services owners I work with who struggle most with their finances aren’t the ones who don’t care. They’re the ones who got busy, trusted a process they didn’t fully understand, and found out six months later that something had been quietly wrong the whole time.
## What the Fix Actually Looks Like
After the initial failure, I didn’t abandon the approach. I rebuilt it — using a second AI tool to help me re-prompt the first one, stress-testing the output against real edge cases, and adding a structured review layer before anything gets used downstream.
That process — design, test, verify, iterate — is exactly what building a functional finance operation looks like. Not glamorous. Not instant. But reliable.
For a professional services firm, that translates to a few concrete things: a clean chart of accounts that reflects how your business actually works, a monthly close process with real review checkpoints, financial reports that tell you something useful instead of just confirming that transactions happened, and someone who knows your business well enough to catch what the tools miss.
That’s what a fractional CFO builds. Not a software subscription. Not a set-it-and-forget-it dashboard. A system with judgment built into it.
## The Part Where I Make the Connection Explicit
I’m sharing this because I think the AI hype cycle and the “just outsource your bookkeeping” advice share the same flaw: both assume that putting a tool in place is the same as solving the problem.
It isn’t. The tool is only as good as the process around it — and the process is only as good as the person who understands what can go wrong.
If you’re a professional services operator who’s ready to stop being your own CFO, the first question isn’t which software to use. It’s whether your financial infrastructure is solid enough that any tool — human or AI — can actually do its job.
That’s usually the most valuable conversation to have. And it doesn’t cost anything to start it.
Enjoyed this article? Consider Sharing
Author
