To build an AI strategy, start from the business, not the technology: define the problems worth solving, prioritise AI use cases by value and readiness, and turn the best ones into a staged roadmap with named owners, budgets and review points. A strategy that begins with "we need to use AI" tends to produce scattered pilots. One that begins with "we need to cut response times by half" gives AI a job to do and a way to measure it.
Step 1: define the business problem
List the three to five business objectives that matter most over the next two years, such as growing revenue in a segment, reducing cost to serve, or improving quality. For each, ask:
- Where is work slow, repetitive, error-prone or dependent on scarce expertise?
- Where do we have data or documents that people struggle to use?
- Where are customers or staff waiting?
These are your candidate problems. Write each as an outcome with a measure: "Reduce average time to answer customer emails from 20 hours to 4 hours."
Our guide to aligning AI strategy with company goals goes deeper on this step.
Step 2: prioritise use cases and readiness
For each problem, list possible AI use cases, then score them on two dimensions:
| Dimension | Questions |
|---|---|
| Value | How much does solving this move a business objective? How many people or customers does it affect? |
| Readiness | Do we have the data? Is the process stable? Is there an owner? What are the risks (data protection, accuracy, bias, security)? |
Plot them on a simple 2 by 2 grid:
- High value, high readiness: pilot first
- High value, low readiness: invest in data and process, then pilot
- Low value, high readiness: quick wins only if cheap
- Low value, low readiness: drop
Step 3: build a staged roadmap with ownership
Turn the top use cases into a roadmap with three stages, each with an owner, budget and a decision point at the end.
A worked roadmap, with assumptions
A fictional professional services firm with 300 staff chooses two use cases: drafting first responses to customer emails, and summarising long contracts for review.
| Stage | Months | What happens | Owner | Decision at the end |
|---|---|---|---|---|
| 1. Prepare | 0 to 3 | Baseline current response times and review hours; agree data and privacy rules; select tools; train 20 users | COO, with IT and Legal | Proceed to pilot if data rules and tools are approved |
| 2. Pilot | 3 to 9 | Two pilots with one team each; human review of all outputs; measure time, quality and errors weekly | Head of Client Service; Head of Legal | Scale, adjust or stop each pilot, based on results against baseline |
| 3. Scale | 9 to 18 | Roll out successful pilots; update processes and training; add the next use case | COO | Annual review of the AI portfolio with the board |
Assumptions, stated openly:
- Human review stays in place throughout; the aim is to speed up drafting, not to remove judgement.
- Time saved is estimated from the pilot, not assumed in advance.
- Costs include licences, integration, training and staff time, not just software.
- Either pilot may be stopped if quality or risk measures are not met.
Writing assumptions down makes the strategy easier to challenge and to update.
Governance and people
Two things make or break an AI strategy after the roadmap:
- Governance: clear rules on data use, human oversight, accuracy checks and who approves new use cases.
- People: training, clear communication about how roles will change, and involving the teams whose work is affected.
Putting it on paper
Once you have the problems, priorities and roadmap, capture them in a short document. Our AI strategy document template shows what to include. When you are ready to take it to leadership, see how to present an AI strategy to your board.
Related reading
- AI strategy subject guide: building, documenting, measuring and presenting an AI strategy
- AI use-case and business-case pack: a worked first-year business case with a downside case
- What does an AI strategy include?
- How to get buy-in for your AI strategy proposal
- How to measure AI ROI and strategy outcomes
If you want to work through this process for your own organisation with a tutor and peers, our AI strategy course for executives covers vision, use-case prioritisation, investment and ROI, governance and the executive roadmap, with no coding required. For a broader introduction, see Strategic Application of AI in Business.
The bottom line
Start from business problems, prioritise by value and readiness, and build a staged roadmap with owners and decision points. State your assumptions, and be willing to stop what does not work.