The 90-day AI plan: what to show after one quarter

The CEO says the company needs an AI strategy. Ninety days later, someone will ask what came of it. This is the plan for having a real answer.

"We need an AI strategy" is a sentence that produces slide decks. AI maturity models, capability heatmaps, vendor landscapes. A quarter later the company owns forty slides and zero moved metrics, and the next budget conversation gets harder.

A quarter is enough for three deliverables that survive contact with a board. Two pages of truth, one shortlist, and two or three pilots with numbers. Here is the sequence.

Weeks 1–2
The two-page current-state memo

Before anyone names a use case, write down what is actually there. Four sections, two pages, no more:

• Systems. What runs the business (ERP, CRM, portal, ticketing) and how they connect, or fail to.
• Data. The honest state of what is inside them. "The CRM is three years of half-filled fields" is a useful sentence; write it if it is true.
• Frictions with money attached. Three to five, specific: quotes take two days, order entry is manual, small customers cost more than they bring, the same report gets rebuilt weekly.
• People. Who could own a pilot next to their day job, named.

The memo reads in ten minutes and does more work than any maturity assessment. Every later decision refers back to it, and it costs two weeks of conversations, not a consulting engagement.

Weeks 3–4
The shortlist

Collect candidate use cases from the teams, from the frictions in the memo, and from documented cases in your industry. Then filter hard, on two questions per candidate: does it move a metric the business cares about, and does the memo say you are ready to run it.

The filter is brutal by design. In one enterprise AI program I worked on, around 200 collected ideas became 15 funded lighthouse pilots. At mid-market scale the same filter typically leaves two or three. That is the right number for one quarter.

Weeks 5–12
Run the pilots

Favor pilots that sit on systems you already run: CRM copilots, inbound triage, document drafts, portal self-service. They start in days and produce numbers inside the quarter. Aerotech ran its sales copilot inside an existing CRM and still needed 3 months for full adoption; that is what realistic looks like when the tool is Tier 1.

Two rules keep the quarter honest. Measure the baseline in week one, because a pilot without a baseline can only produce anecdotes. And keep a kill log: a candidate you stopped, with the reason, is evidence of judgment, and boards read it that way.

Day 90
What you show

One page again:

• Two or three metrics, each with baseline, current value, and trend.
• What each pilot costs and what it returns at current trajectory.
• What you killed and why.
• The next two candidates from the shortlist, already bracketed.

That page is the AI strategy. It states where the business bleeds, what was tested against real numbers, and what earns investment next. The maturity heatmap never has to exist.

Frequently asked questions