The first 90 days of an AI adoption plan.

On day 90, one capability is in daily use by the roles that own the work. Your team has evidence that it holds on real work: the cases it ran, what the reviewer accepted and what it sent to a person. And you have a ranked list of what comes next, each item with an owner and its inputs named. That is the first step toward 10X what your team can do, and the second capability starts from a map you have already drawn.

Many organizations already have the AI. The subscriptions are paid for, a training session has run and a few people get excellent results. It is a familiar pattern: everyone has the AI and one person gets the result. The first 90 days turn that one person’s results into a capability the whole team runs.

Ninety days keeps the plan specific and still leaves time to prove the work on real cases before anyone calls it done. The plan runs in seven steps, week by week. A completed roadmap for an invented organization and a blank one for yours follow the steps.

The plan in seven steps.

1. Name an owner.Week 1. One person accountable for the outcome.
2. Choose the outcome and map the work by role.Weeks 1 to 2. What the team will be able to do, and the work that gets it there.
3. Select one pilot.Week 3. The piece of work to prove first.
4. Establish a baseline.Weeks 3 to 4. How the work runs today, measured by your team.
5. Build and test on real work.Weeks 4 to 8. Capture the method, write the tests, run real cases.
6. Roll out.Weeks 9 to 11. Daily use by the roles that own the work.
7. Decide whether to expand.Weeks 12 to 13. Expand, adjust or stop, on the evidence.

The illustration throughout is an invented 120-person regional distributor. Its order desk starts each morning with a queue of held orders: credit holds, incomplete delivery addresses and stock that is not where the system says. Clearing them passes through the order desk coordinator, the credit controller, the warehouse supervisor and the customer service lead.

Week 1: name an owner.

The owner is accountable for the outcome the business wants. Choose the leader whose team does the work: an operations lead, a controller, a firm administrator, a head of client success. The owner sets the scope, runs a short weekly check and makes the day-90 decision. A CEO or managing partner sponsors the plan, clears time for the people involved and hears the decision.

In the illustration, the CEO names the operations lead, who already owns order fulfillment.

Weeks 1 to 2: choose the outcome and map the work by role.

Write the outcome as something the team will be able to do every time. “Every customer knows where their order stands before noon” is an outcome. “Use AI in order management” is a topic.

Then map the work that serves the outcome. Write each piece of work once, as a shared record: what starts it, what finishes it, the roles it passes through, the handoffs and the decisions a person takes. An AI org chart should show the work between teams walks through one record and its view for each role. Draw the map with the people who do the work; build your AI plan with the people who do the work gives five questions to ask them. For intake, billing and client service at a law firm, the law-firm adoption plan applies the same map to the business side of the firm.

The distributor maps the four pieces of work that leave a customer waiting for an answer: held orders, returns authorization, backorder notices and delivery-date changes.

Week 3: select one pilot.

One pilot chosen well teaches more than four started at once. Score each candidate on five questions.

• Does it recur?Daily or weekly work gives enough cases to test and enough use to build the habit.
• Does it cross a handoff?Work that passes between roles is where prepared work pays back fastest.
• Are the inputs available?The records exist, and the team is permitted to use them in its AI workspace.
• Is a miss caught?A person reviews the output before anything consequential happens.
• Will the role owner run it?Someone who does the work wants this one first.

The readiness checklist covers inputs, review and approvals in more detail. The distributor chooses held orders: they arrive every morning, cross four roles and already have a reviewer at every consequential step. The pilot gets its own line of the outcome: every held order is cleared, or explained to the customer, before noon. Returns authorization goes second on the list.

Weeks 3 to 4: establish a baseline.

Before anything changes, measure how the work runs today on recent real cases. Use five measures.

1. Elapsed time.From the trigger to the finished output.
2. Hands-on effort.The time people spend working on it, by role.
3. Review.The time the reviewer spends checking it, and how often it goes back.
4. Rework.How often a finished output is corrected after it has moved on.
5. Output quality.The reviewer’s judgment against an agreed standard: complete, correct and ready for the next role.

Your team takes these measurements on its own work, and every figure in the plan comes from your records. The same five measures, taken the same way, return in weeks 12 and 13. In the illustration, the order desk coordinator records them across recent held orders and the operations lead accepts the baseline.

Weeks 4 to 8: build and test on real work.

Capture the method from the people who do it. Walk through a routine case and an awkward one, then write down the steps, the inputs, the decision rules and the exceptions. SOP exception capture finds the cases the written procedure leaves out: the customer with a standing credit arrangement, the address that is incomplete but deliverable.

Write the acceptance tests before the first run. Each test names an input situation and the expected behavior, including missing, contradictory and ambiguous inputs. Pilot acceptance-test preparation builds that set from the captured method. Then run the workflow on held-back real cases and have the reviewer record each one: what was expected, what came back and what they decided. What we test before we call an AI workflow ready shows the full test record.

In the illustration, the workflow prepares the morning’s exception list with the evidence for each hold, a proposed next step and a draft customer note. The credit controller decides every credit release. The warehouse supervisor confirms any substitution. The customer service lead approves and sends each customer message.

Weeks 9 to 11: roll out to the roles that own the work.

A workflow that passes its tests becomes a capability when the team uses it every day. Introduce it to each role with a one-page job aid: when to run it, what to supply, how to review the output and what to do with an exception. Sit with each person through their first real uses. Name who answers questions, and keep a log of exceptions and suggestions that the owner reads at the weekly check. The rollout checklist lists every item.

From week 9, the distributor’s order desk coordinator runs the workflow each morning, and the credit controller, the warehouse supervisor and the customer service lead work from the same prepared list.

Weeks 12 to 13: the day-90 decision.

Retake the five baseline measures the same way, on the same kind of case. Read them with the test record and the rollout log, then make one of three decisions.

Stopping is a sound decision. The map, the captured method and the baseline stay with the business, and the next pilot starts better informed. Here is the distributor’s record.

Illustrative example with invented data. The distributor’s day-90 decision.

A completed 90-day roadmap.

The whole plan for the distributor, on one page.

Illustrative example with invented data. An invented 120-person regional distributor.

A blank roadmap to copy.

Copy this into a document your team shares. Answer each row’s questions before its weeks begin, and replace each question with the answer.

Your platforms, your pace, your people.

The capability runs in the AI your organization already pays for. Fully Powered works across ChatGPT, Claude, Copilot and Gemini, and every step above happens in your workspace, with your records and your approvals. The owner sets the pace. The people who do the work shape the method and keep the decisions that are theirs. We help your organization strategically plan and migrate to AI without losing control of how the work gets done.

This is AI capability planning in practice: start from the outcome, map the work that crosses teams and land each capability on the roles that do it. After day 90 the same discipline becomes AI capability management. Each capability on the ranked list gets its own owner, baseline, test record and decision. That is how you 10X the AI you already pay for: one proven capability at a time.

Start your first 90 days.

Bring the outcome you want on day 90 and the work you think crosses teams.