AI Strategy and a 90-Day Roadmap

An AI strategy is a short plan naming the business problems AI will solve, who owns each one and how you will measure results. A 90-day roadmap sequences it: choose in month one, pilot in month two, measure and decide in month three.
This guide covers the parts of the plan, the difference between a strategy and a roadmap, and a step-by-step first 90 days. For help running it, see the consulting page or my fractional Chief AI Officer service.
What is an AI strategy?
An AI strategy is the plan for how a company will use AI to meet its business goals. It answers three questions: what do we want AI to do for the business, who is responsible, and how will we know it worked.
Microsoft’s AI strategy guide names five building blocks, and the list is a useful checklist because each one can be answered in a sentence.
| Part | The question to answer in one sentence |
|---|---|
| Executive vision and alignment | Which business goal is AI serving this year, and who at the top backs it? |
| Use case prioritization | Which one or two problems do we start with, and why those? |
| Data and infrastructure | What data and tools do those use cases need, and do we have them? |
| Talent and skills | Who builds, who uses, and what training do they need? |
| Governance and ROI metrics | What rules apply, and what number will tell us it paid off? |
Source: Microsoft, AI Strategy for Business Transformation, read November 11, 2026.
What is the difference between an AI strategy and an AI roadmap?
The strategy is the what and the why. The roadmap is the when and the who. A strategy says “we will cut the time our team spends on customer replies.” A roadmap says which team starts in which month, who owns it, what is measured and when leadership decides to continue or stop.
Skipping the strategy gives you a list of tools with no goal. Skipping the roadmap gives you a goal with no next step. The 90-day plan below covers both for your first cycle.
Why does a company need an AI strategy?
Because people are already using AI whether or not leadership has a plan. Three data points show the gap.
- Leaders know it matters but do not feel ready. In Microsoft and LinkedIn’s 2024 Work Trend Index, 79% of leaders agreed their company needs to adopt AI to stay competitive. Yet 60% worried their organization’s leadership lacks a plan and vision to implement it.
- Staff move first. In the same report, 78% of AI users were bringing their own AI tools to work, and the figure was 80% at small and medium-sized companies.
- Pilots rarely pay off at scale. IBM’s 2025 CEO study of 2,000 CEOs found that only 25% of AI initiatives had delivered the expected return over the last few years, and only 16% had scaled across the enterprise. Half of the CEOs said rapid investment had left them with disconnected, piecemeal technology.
Sources: Microsoft and LinkedIn, 2024 Work Trend Index, published May 8, 2024; IBM Institute for Business Value, 2025 CEO Study, published May 6, 2025.
A plan does not guarantee a return. It does stop the three patterns above: tools bought with no goal, staff using unapproved tools, and pilots that never reach daily use.
How do you build an AI roadmap in 90 days?
Split it into three 30-day steps, and end each with one decision. Published 90-day plans from consultancies and software vendors share this shape, with different names for the steps. One example, from Alice Labs (updated September 16, 2026), describes it as “a cadence of three decisions” with one milestone, one deliverable and one checkpoint per phase. It is a vendor page, so treat it as one view, not a benchmark.
The roadmap below follows the same cadence. It is built from that common shape and from the research cited in this guide, not from a controlled study.
| Days | Step | What you produce | Decision at the end |
|---|---|---|---|
| 1 to 30 | Choose | A one-page pilot brief: goal, one or two use cases, owner, baseline, rules | Does the sponsor approve the pilot? |
| 31 to 60 | Pilot | A working pilot used by a small group on real work, with results against the baseline | Continue, change course or stop? |
| 61 to 90 | Measure and decide | A short results report and a 12-month plan | What do we scale, and what do we fund? |
Days 1 to 30: How do you choose where to start?
Start with a business goal, not a tool. In the first month you decide what AI should improve, pick one or two places to start and write down the rules. Do these six things.
- Pick one goal. Choose a number the business already tracks, such as hours spent on a weekly report, days to answer a customer or cost per lead. This is what the roadmap serves.
- Walk through the work. List the tasks that eat the most time in the team tied to that number. Ask the people who do them, not only their managers.
- Find out what people already use. Run a short, no-blame survey of which AI tools staff use and for what. Given the 78% figure above, expect an answer.
- Score the candidates. Rate each task on value, ease, data access and risk (see the next section). Keep the top one or two.
- Name one owner. One person is accountable for the pilot and has time set aside for it.
- Write a one-page rule sheet. What data may not be pasted into AI tools, who approves new tools, and what must be checked by a human before it goes out.
Record the starting number before you change anything. Without a baseline you cannot say, at day 90, whether anything improved. An AI readiness assessment is a quick way to check where you stand on data, rules and people before the pilot.
The exit test for month one: the sponsor signs a one-page pilot brief with the goal, the owner, the baseline and what counts as success.
How do you pick the first AI use cases?
Pick tasks that are frequent, easy to check and low in risk. A first pilot should be boring on purpose. For ideas by department, see AI use cases for business. The aim is to show that your team can run an AI project from start to finish.
| Test | Good first use case | Poor first use case |
|---|---|---|
| Value | Happens daily or weekly and costs real hours | Happens twice a year |
| Ease | A person can check the output in minutes | Output is hard to judge |
| Data | The inputs already exist in tools you own | Needs months of data cleanup |
| Risk | Internal, with a person reviewing the result | Customer-facing with no review |
RAND Corporation interviewed 65 experienced data scientists and engineers in 2024 about why AI projects fail. The most common root cause they gave was that stakeholders misunderstand or miscommunicate what problem AI is meant to solve. Source: RAND, The Root Causes of Failure for Artificial Intelligence Projects, published August 13, 2024. That is why step 2 above puts the people who do the work in the room before anyone picks a tool.
Days 31 to 60: How do you run the pilot?
Keep it small, real and measured. A pilot is a short test of one use case with a handful of people, doing their normal work, with a person reviewing the output.
- Use real work. A demo on sample data tells you little. Run the tool on this week’s actual tasks.
- Keep a human check. Anything that leaves the building, such as an email to a customer, is reviewed first.
- Meet briefly and often. A fifteen-minute check-in twice a week surfaces problems while they are cheap to fix. Expect the first week to feel slower while people learn the tool.
- Train as you go. Show the pilot group how to write good instructions, how to check for wrong answers and when not to use the tool. My AI workshops cover this live for teams.
- Track against the baseline. Measure the same number you recorded in month one, plus the tool cost and the time people spend on the pilot.
I treat ideas the same way: as assumptions, not conclusions. Test them instead of jumping ahead and letting your ego decide you know best.
At day 60, make one of three calls: continue as planned, change the approach, or stop. Alice Labs suggests treating a miss of more than 30% against the target as the point to choose explicitly. Stopping a pilot that does not work is a good outcome: it saved you from scaling the wrong thing.
Days 61 to 90: How do you measure results and decide what to scale?
Write the results down in one page, then decide. A simple report has five lines: the goal, the baseline, the result, the total cost (tools plus staff time) and what you learned.
Then answer three questions:
- Did it beat the baseline by enough to matter? If yes, how much time or money is that per year? The AI ROI guide shows how to work that out.
- What would it take to roll out to the full team? Training, access, data, review steps and cost.
- What is next? Pick the second use case from your month-one list, and set the next 12 months in order.
Review the rule sheet at the same time. By now you know which rules people followed and which they worked around. Fix the ones that did not fit real work.
This is where a 90-day roadmap becomes a company plan: you leave with a measured result, a funded next step and a short list of what comes after. To take a pilot to the whole company, see the AI adoption guide, and for the wider change across the company, see AI transformation.
What mistakes sink an AI strategy?
Watch for these six.
- Starting with a tool. Buying first and finding a use later is the quickest route to disconnected technology, which half of the CEOs in IBM’s study reported.
- No owner. If everyone is responsible, nobody is.
- No baseline. You cannot show a gain you did not measure.
- Too many pilots at once. Five loose tests teach you less than one well-run one.
- No rules. Staff will use AI regardless, so the rules decide whether it is done safely. Even a one-page sheet beats none, and the AI governance guide covers the next level.
- Treating day 90 as the end. Goals, tools and rules change, so the plan needs a review. Put one on the calendar every quarter.
Who should own the AI strategy?
One senior person with authority across teams, and enough time to do it. In a small company that is often the CEO or COO. As AI spreads across departments, many companies add a dedicated leader.
That leader is often called a Chief AI Officer. Companies that do not need a full-time one sometimes use a part-time or fractional Chief AI Officer. These guides cover the role in more detail:
- What a Chief AI Officer does
- When to hire a fractional Chief AI Officer
- Fractional vs full-time Chief AI Officer
- What a fractional Chief AI Officer costs
What should you do next?
Write the one-page pilot brief this week. Choose the goal, the number you will track, the owner and the first use case. To see what paid help on this should deliver, read AI strategy consulting. If you want a second set of eyes on the brief or help running the pilot, my fractional Chief AI Officer service owns the plan, the policy, the training and the shared systems.
Sources
- Microsoft, AI Strategy for Business Transformation, read November 11, 2026.
- Microsoft and LinkedIn, 2024 Work Trend Index: AI at Work Is Here. Now Comes the Hard Part, published May 8, 2024.
- IBM Institute for Business Value, 2025 CEO Study, published May 6, 2025.
- RAND Corporation, The Root Causes of Failure for Artificial Intelligence Projects and How They Can Succeed, published August 13, 2024.
- Alice Labs, AI Strategy Roadmap: The 30/60/90 Day Plan, updated September 16, 2026.


