Many people picture an AI strategy as a thick document. In practice, the strategies that actually get used often fit on one page. Below is a worked example for a fictional 200-person professional services firm. It is not a template to copy word for word, but it shows the level of specificity that makes a strategy useful.
The example: "Northbridge Advisory" AI strategy, year one
Purpose. Give our consultants back time on low-value drafting and admin so they spend more of their week with clients, while protecting client confidentiality.
Where we will focus (in order).
- Proposal and report first drafts.
- Meeting notes, actions and client follow-ups.
- Research summaries for new client briefings.
What we will not do this year. Build custom AI models; use AI to make client recommendations without senior review; automate any client-facing communication without human approval.
Approved tools. Business versions of one general AI assistant and the AI features in our existing office suite. No other AI tools with client data.
Data rules. Never enter client names, financials or personal data into non-approved tools. Approved tools only, with confidential documents handled per our data policy.
People. Every consultant completes hands-on training on the three focus tasks by end of Q1. One AI champion per practice area. Partners trained first.
Owner. Operations Director, reporting quarterly to the leadership team.
How we will measure. Hours per proposal; weekly active users; rework rate on AI-assisted reports; client satisfaction on affected engagements.
Milestones. Q1: training complete, pilot in two practice areas. Q2: standard workflows documented, rollout to all practices. Q3 to Q4: review results, choose next three use cases.
Why this works
- It names priorities in order. Teams know what to work on first.
- It says what not to do. This is as useful as what to do, and reduces risk.
- It is specific about data. Vague "use AI responsibly" statements change nobody's behaviour.
- It has an owner and a review rhythm. Someone is accountable and progress is checked.
- It measures outcomes, not activity. Hours saved and quality, not licences.
How to write your own
Start with the purpose in one sentence. Then choose your first two or three use cases by asking teams which recurring tasks waste the most skilled time. Write the data rules in plain English, with your data protection lead. Name the owner. Pick three or four measures, and set quarterly milestones.
If you want a fuller document for a board or a larger organisation, our guide to what goes in an AI strategy document shows how to expand each section. But start with one page: if you cannot write it on one page, you probably do not yet know what you are doing.
The step before writing
A strategy like this is much easier to write once you have used AI tools on real work yourself. You will know which use cases are realistic and which are hype. On our Leading AI Innovation & Organisational Transformation course, participants build and present a plan like this for their own team as the course assessment.
Frequently asked questions
Is a one-page AI strategy enough for a large company?
As a summary, yes. Large organisations will need supporting detail on governance, budgets and integration, but the one-page version keeps everyone aligned on what matters.
How often should we update the strategy?
Review it quarterly and rewrite it annually. AI capabilities change fast enough that a strategy older than a year is usually out of date.
Who should see the AI strategy?
Everyone affected by it. A strategy staff never see cannot shape their behaviour, especially the data rules.