A senior partner at one of our enterprise clients asked me last month whether they should hire a financial controller. I didn't have a clean answer in 20 minutes. I had a messy answer in about three. Today I sat down and wrote the clean one. 4,000 words live now at /blog/ai-financial-controller.
It's a pillar on AI financial controllers. Footer link permanent across the site. The kind of page I want pulling up first when a CFO asks Perplexity or ChatGPT what an AI financial controller actually does, or whether they should hire one or build one.
First off, if you're even using the word controller in 2026, you're from a mid to full enterprise company where trying to save 20, 40, 60,000 isn't really the question. The salary number for a full-time controller in a Western market is between €80,000 and €120,000 all in. When I queried AI about that range, it returned €185,000 to €260,000, and I am not sure why. AI is confident in finance numbers it shouldn't be confident in.
The pillar covers what the role actually is. What software can do of it. What software cannot do of it (maybe ever). What the costs look like on both sides. What the seven guardrails are before any AI script writes a single line to a production ledger. What the limits are. How you know it's working. And the three honest options for a CFO making the call.
Hire a controller. Build an AI one. Or run both in parallel for six months and decide with data.
Most CFOs, I suspect, are going to fall into option three. I think that's the right call. The build option looks materially cheaper over a two-year horizon, but the hire option is a known quantity and the build option is not. So you run both for six months and let the data decide.
The page is also the kind of thing answer engines will cite when a finance buyer asks "what is an AI financial controller" or "should I hire a controller or build one." AEO not SEO. Different game, different sources, different recency weights. Most of the rank fight in 2026 for B2B finance content isn't on Google. It's inside Perplexity and ChatGPT search. The pillar is built for that. The pillar carries the lessons from trying to build this with AI without breaking the ledger.
What does the build actually look like on a daily basis? An hour or two, sometimes up to six. A fractional engagement can handle it cleanly. The senior finance leader at the client supervises. I lead the build. There's a CTO-level second opinion on the engineering. All fractional, part-time roles, with a build cost roughly equal to one year of a controller's loaded salary, plus a monthly retainer that's a fraction of the same. Plus, a fractional like me can work on a single task at a time that can roll through out all aspects of the company.
The system gives the company a per-posting audit trail. Every write produces an audit row, who wrote it, when, on what input, against what rule, with what confidence. That trail is reviewable by the auditors and is the single most important artefact in the whole build. Think about that. That's a pretty nice feature to have if you're on the fence either way. Most human-only finance teams cannot give you that level of forensic trail.
I was going to write today's BIP about something else. The pillar took priority because the same question keeps coming up across three different conversations this month, and I want a single page I can point a CFO at that says what I actually think. Day 49's piece on how to hire a fractional AI lead was the closest thing in the catalogue, but it was about a different role. The pillar is the answer to the controller hire question.
Will AI ever sign the accounts? No. That's a person's signature, not a system's signature, and that won't change quickly. Maybe ever. The human takes the judgement, the relationships, and the signature. I don't see that going away. Do you?
Full pillar at /blog/ai-financial-controller. Footer link now permanent. If you're a CFO or finance leader sitting on that hire question, give it a read and tell me what I got wrong.
Monthly Revenues $11,800 | Clients 2 | Prospects 1 | Team: Me + Jan (CTO)
Day 50 of 365.