AI is not an IT project. Put it next to headcount.

AI is not an IT project. Put it next to headcount.

Most shops hand AI to IT, wait for a deck, and move on. That is how it stays a side quest. It belongs in the annual plan with people, capex, and revenue. Planning, finance, and strategy own that, not a skunkworks slide.

Do the data inventory before you pick a budget. Quality: dates that match, no silent duplicates, fields that are actually filled. Access: can the people who need it get it, or is it locked in a department. Ownership: HIPAA, GDPR, who can say yes. Where it lives, and whether anything talks to anything else. Weakness in one of those sinks the rest.

Two or three use cases. Price the whole thing.

For each goal: is there a slow or inconsistent decision, clean data, and a number you can watch? If not, skip it. Two or three is plenty. More than that dies in May.

Budget the parts people forget. Paid tools, not free tabs that train on your files. Data prep, which is most of the hours. Upkeep: a field you keep clean has a real annual cost. Then pick a leading number and a lagging one per use case. Hours saved. Completeness. Validated outputs this week. Faith is not a KPI.

Price the risks in the same meeting

Hallucinations on numbers that go to the board: a person reads them. Default training on your inputs: turn it off or do not use the tool. Cutoff dates: if you need this week, do not use a model that stopped last year. Access creep: everyone with a login to internal data is an audit line.

Review it quarterly. Models drift. Data drifts. Strategy drifts. The inventory is the first job of the next cycle. Without it you are budgeting on a vibe.