Six months of testing AI inside a live corporate finance function. Not a demo. Not a vendor pitch. Real month-ends. Real deadlines. Real compliance constraints. Nothing invented.
I have been testing AI in a real corporate finance job for the past six months. Not in a sandbox. Not in a demo. During actual month-end closes, CAPEX reviews, and board preparation — inside a listed pharmaceutical company with real governance constraints and real deadlines.
Nobody asked me to document this. I started because I could not find anyone else doing it honestly. Everyone sharing AI content in finance is either selling a tool, running a consultancy, or demonstrating workflows that would never survive inside a real corporate environment. The demos look clean. The reality is different.
This newsletter is the reality.
Month-end commentary. CAPEX variance analysis. CFO challenge preparation. Board narrative drafting. The tasks that take the most time and carry the most political sensitivity inside a senior finance role.
Every test used anonymised data. No company names. No project identifiers. No absolute figures that could identify the business. The Safe Input Framework I have built and shared on the website is exactly what I used — and it worked.
It separated two variance types I had conflated. Project delay underspend versus close-out savings release. Those are different stories. The Finance Director notices when you combine them. The AI caught it before I did.
AI identified the structural distinction between variance types that I had blurred under deadline pressure. It produced board-ready narrative in 22 minutes — a task that typically takes me two hours from raw data to final commentary.
Then it assumed two delayed projects would recover in H2. They will not. That assumption — stated confidently, formatted neatly, looking completely plausible — would have gone straight to the board pack if I had not read it carefully.
AI has no visibility into the conversations behind the numbers. It does not know that Project A is delayed because of a strategic decision that will not reverse. It reasons with data. It cannot reason with context. That distinction is everything in senior finance work.
I caught it. Fixed it. Used the corrected version. Net result: 60 minutes saved and one near-miss avoided. That is a good month-end.
AI accelerates the work. It cannot replace the understanding. That is not a limitation to work around. It is a feature to design for.
I gave AI an anonymised version of my CAPEX portfolio and asked it to generate the questions my CFO would most likely ask in the review meeting. It produced fifteen questions. I ranked them by likelihood, took the top five into the meeting.
Three of the five came up exactly as written. A fourth was a valid version of what was asked. The fifth was not asked that day — but it should have been, and I was ready for it if it had been.
CFO question anticipation is the highest-value use case I have found for AI in senior finance work. It forces you to stress-test your own analysis before you present it. The AI asks the questions your stakeholders will ask. Your job is to answer them before the meeting — not during it.
The prompt that produced this result is in Category D of the free prompt library. I have used it every month since.
The finance professionals who will use AI well are not the ones who use it most. They are the ones who use it most safely — with the right input protocol, the right validation habit, and the right understanding of where the tool's judgment ends and theirs begins.
Nobody is teaching that combination in a way that works inside a real corporate environment. That is what this newsletter is for.
Safe input first. Always. Anonymise before you prompt. Replace company names, project identifiers, and absolute figures. Never paste live data into any AI tool on any device.
Your judgment is the last layer. Always. AI produces structure and narrative. You correct what it cannot know — the context, the politics, the six months of conversations behind the numbers. Review everything. Correct what is wrong. Then use it.
CAPEX with AI — A Real Investment Case. A £20M capital project. AI applied at every stage — from investment narrative to variance analysis to board presentation. What it produced, what it missed, and what the governance trail looked like. The most detailed finance AI walkthrough I have published.
The experiment continues. This contract ends in July and the next senior role search starts immediately — with AI used at every stage of the process. The applications, the company research, the interview preparation. All of it documented, all of it shared here. Then the onboarding and the first month-end in a new environment. Real time. Nothing edited out.
If you are reading this — you are already ahead of the gap.
— Petar · The AI Enabled Finance
Subscribe free and get 25 AI finance prompts plus access to the Members Hub — every issue delivered before it goes public on LinkedIn.
No spam. Unsubscribe anytime. No company data ever required.