What a fractional Chief AI Officer does in the first 90 days

Most descriptions of the Chief AI Officer role stay abstract. Owns the strategy, sets the direction, drives adoption. None of that tells you what lands on your desk in month one, which is the only question that matters when you are paying a retainer. So here is the concrete version: the documents, decisions, and meetings a fractional CAIO should produce in the first 90 days, in order.

If you are still deciding whether the role is needed at all, start with Do you really need a Chief AI Officer? This article assumes you have answered yes and want to know what you are buying.

Days 1 to 30: find out what is actually true

The first month is investigation, not action. Executive transition research is consistent on this point: the leaders who last are the ones who resist the urge to launch things in week two. A fractional CAIO who arrives with a solution before seeing your numbers is selling the solution, not the judgment.

Four artefacts should exist by day 30.

The tool and spend inventory. Every AI tool in the company, who bought it, what it costs, what data it touches, and whether anyone still uses it. In most mid-market companies this list has never been written down, because departments bought independently. The inventory almost always surfaces overlap and abandoned subscriptions, and the recovered spend is often the first visible return on the retainer.

The decision list. Every AI decision currently open or postponed: the vendor proposal sitting in someone’s inbox, the pilot with no owner, the build-or-buy question nobody has called. Each entry gets a dollar consequence and a current owner, even if the owner is “nobody”. This list becomes the working agenda for everything that follows.

The opportunity map. The inventory and decision list describe what exists. This artefact describes what could. It comes from structured interviews with department heads, walking through where hours are spent, where errors cost money, and where customers wait, then marking the points where AI plausibly changes the number. The output is a candidate list with rough sizing, not commitments. Every company has more candidates than capacity, which is exactly why the list must be written down before anything is chosen: unwritten, the choice drifts toward whatever is technically interesting rather than what moves revenue, cost, or risk.

The state-of-AI summary. One document, written for the leadership team, saying what the company actually has, spends, and risks today. It is a baseline rather than a vision. If your board has been asking about AI strategy, this document is the first credible thing you can put in front of them.

Days 31 to 60: choose, and start one thing

Month two merges the decision list and the opportunity map into one ranked agenda, sized by effect on your numbers rather than by enthusiasm. Most items get one of three verdicts: decide now, schedule for a named quarter, or kill. The killing matters as much as the choosing. A portfolio of nine half-alive pilots is worth less than two funded ones with owners, and an opportunity that cannot beat the items above it on the list is a distraction however clever it sounds.

One initiative starts in this window: the first scoped win, drawn from the top of the ranked agenda. It is chosen for a specific property, that it can succeed visibly within a quarter and be measured openly. If your earlier AI projects failed and the team is sceptical, this is the deliberate reset. Small in scope does not mean small in effect. The point is to re-establish, with evidence, that AI can work inside your company on your data with your people.

What should not start in this window: platform commitments, reorganisations, hiring plans. Any of those made before day 60 are being made on someone else’s pattern-matching rather than your facts.

Days 61 to 90: write down the rules

Month three produces the governance baseline. This is the least glamorous deliverable and the one most likely to be requested by someone outside the company, because enterprise customers now send AI-specific questionnaires and boards now ask AI-specific questions. The baseline answers four things in writing: who approves new AI initiatives, how competing requests get prioritised, who is accountable for anything running in production, and how a problem escalates when one appears. We cover what customers and auditors actually ask in The AI governance baseline.

The board cadence also gets set here: what gets reported, how often, and in what form. A one-page quarterly position beats a slide deck of experiments every time.

What you hold at day 90

At the end of the first quarter you should be able to open a folder and find the tool inventory, the opportunity map, the ranked decision agenda with verdicts, one scoped initiative in flight with a metric attached, the governance baseline, and a board-ready statement of position. Six artefacts. If a fractional CAIO cannot describe their first 90 days at this level of specificity before you sign, that tells you something useful. We list the other questions worth asking in How to evaluate a fractional CAIO before you sign.

The cost side of this comparison, retainer against the $300k to $550k all-in cost of the full-time hire, is covered in Chief AI Officer salary in 2026. The short version: the first 90 days look nearly identical in both models. What differs is what you paid for them.

If you want the first-month baseline built independently before committing to any retainer at all, that is the exact shape of our diagnostic.