A Chief AI Officer's First 90 Days, Done Properly
The first quarter decides whether the seat earns its place. Here is the sequence that works, and the launch that wastes it.
Erin Moore
Fractional Chief AI Officer
A Chief AI Officer's first 90 days should end with a decision set, not a launch. Specifically: an inventory of what you already run, a measured baseline for the processes worth changing, a ranked list of what to fund, and at least one thing publicly stopped. Anything that ships in the first quarter was probably already underway.
The mistake that wastes the window is launching something visible early to prove the hire was justified. It produces a project chosen for speed rather than value, and it spends the political capital the seat needs later for the harder decision — stopping things.
Days 1 to 30: find out what is actually happening
Inventory every AI tool in use, what it costs, who owns it, and what data it touches. Include the unsanctioned ones; finding shadow AI is usually the most revealing part of the exercise and almost always uncovers duplicate spend.
At the same time, interview the people doing the work rather than the people describing it. The gap between the documented process and the actual one is where most automation projects die, and it is only visible in the first month while you are still allowed to ask naive questions.
Deliverable: a register of AI systems and spend, and a shortlist of processes worth examining.
Days 31 to 60: measure before you touch anything
Establish baselines for the shortlisted processes — how long they take, how often they run, what they cost, what the error rate is. This is unglamorous and it is the step that makes every later claim defensible. Without it, you will be arguing about whether things improved rather than showing it. The ROI calculator exists to force these numbers into the open.
Simultaneously, establish the governance minimum: approved tools, prohibited data, who approves new tools. One page. The policy template is a starting point you can adopt in an afternoon rather than a project.
Deliverable: baselines for three to five processes, and a governance page people have actually read.
Days 61 to 90: decide, and stop something
Rank the candidates on value, feasibility and reversibility — the method in how to prioritize AI use cases — and fund the top one or two. Then kill something. There is almost always a tool nobody uses, a pilot nobody finished, or a subscription nobody can defend.
Stopping something in the first quarter matters more than starting something. It establishes that the seat has the authority to end work, which is the authority the role actually depends on. If a Chief AI Officer can only add to the portfolio, the portfolio will only ever grow.
Deliverable: a funded shortlist, a stopped project, and a one-page report to leadership stating the baseline each funded item will be judged against.
What good looks like at day 90
Leadership can answer three questions they could not answer on day one: what are we spending on AI, what is it producing, and what are we doing next quarter. That is the whole job, demonstrated once. The 90-Day AI Playbook walks the same sequence in more detail if you are running it yourself rather than hiring for it.
A public structure for the governance step
The governance minimum in days 31 to 60 does not need inventing. NIST's AI Risk Management Framework organises the work into govern, map, measure and manage, which maps closely onto the inventory, baseline and review sequence above — and gives you an answer when someone asks which standard you followed.
Frequently asked questions
What should a Chief AI Officer do in their first 90 days? Inventory existing AI use and spend, measure baselines for the processes worth changing, establish a one-page governance minimum, rank and fund a shortlist, and stop at least one thing. The output is a decision set, not a launch.
Should a new AI executive ship something quickly? No. Launching early to justify the hire selects a project for speed rather than value and spends credibility that the role needs for stopping work later. Anything shipped in the first quarter was usually already in flight.
Why does stopping a project matter so much? Because it establishes that the seat can end work, not just fund it. Without that authority the AI portfolio only grows, and nothing is ever declared finished or failed.
What if there is no baseline data to measure against? Then creating it is the first deliverable. A process whose current cost nobody can state cannot be improved defensibly, and measuring it is worth doing whether or not you automate anything.
Further reading
Tagged with:
Ready to Automate Your Business?
Let's discuss how AI automation can deliver measurable ROI for your organization in 90 days or sooner.