Eighty-eight percent of enterprise transformation strategies fail to achieve their original ambitions. Not because the strategy was wrong, necessarily - but because no one looked at it again after it was approved. That's a Bain finding, and it should give you pause if you're currently halfway through drafting your firm's AI strategy.
Here's what I've seen happen, more times than I'd like to count: a managing partner or CTO spends six weeks pulling together an AI strategy document. It goes through three rounds of internal review. The board signs it off. Everyone feels good about it. And then it goes into the shared drive, and that's where it dies.
We're producing this because the board asked for it. Once it's done we can get back to the real work.
If that thought has crossed your mind - even briefly - you're not alone. And I'm not here to judge it. The board did ask for it. The pressure is real. But the document you produce in response is either going to change what your firm actually does with AI over the next twelve months, or it's going to serve as proof that you ticked a box. There's not really a middle ground.
I was in a session with a 200-person consulting firm last year - the kind of firm that does genuinely good work, well-regarded in their sector - where the managing partner pulled up the AI strategy on screen and went quiet for a moment. Nobody had opened it since March. It was October. Seven months of AI development, two regulatory updates, and a competitor launching a client-facing AI tool, all of which had happened after the document was written and before anyone had thought to revise it. That's not unusual. That's the norm.
I've helped produce and review enough of these documents across professional services firms - law, consulting, accountancy, financial advisory - that some clear patterns have emerged. The ones that gather dust tend to share certain features. So do the ones that get used. This piece is about the difference.
Let me be direct: this isn't the only way to structure an AI strategy. Different firms at different stages of maturity will weight things differently. But across the documents I've seen work in practice - the ones that changed budgets, shaped pilot selections, and gave leadership teams something to hold themselves accountable to - these seven sections kept showing up.
1. Current state assessment. Where is the firm today? Not a polished narrative - an honest profile across five dimensions: data quality, systems infrastructure, cultural readiness, governance maturity, and skills. If you've done a formal AI readiness assessment, this section populates directly. If you haven't, it forces you to do the work. The temptation is to make it sound better than it is, because the board will read it. Resist that. A strategy built on an optimistic baseline produces a plan that doesn't match reality, and reality wins every time.
2. Opportunity identification. Three to five specific AI applications, prioritised by commercial value and implementation feasibility. Not a list of forty things AI could theoretically do. Forty is not a strategy - it's a research project. Three to five says: we looked at the full picture, we made decisions, and here's what we're backing first. Each one should have enough detail that someone could brief a pilot from it.
3. Readiness gaps. For each of the prioritised opportunities, what specifically needs to be true that isn't true today? If your top priority is automated document review but your document management system stores everything as unstructured PDFs with no consistent metadata, that's a gap. And it's a gap that has a cost and a timeline to close. Be specific. "We need better data" is not a gap analysis. "Our matter management system contains 14 years of client data across three inconsistent taxonomies, and unifying it will require approximately X weeks of work" is.
4. Governance framework. Who decides what gets funded. Who is accountable for delivery. What oversight looks like in practice. This is the section most firms either skip entirely or fill with platitudes about "responsible AI." Neither works. The critical thing: the people named in this section need to know they're in it, agree to what it asks of them, and understand specifically what they're supposed to do. A governance framework that exists on paper but not in anyone's calendar is decoration.
5. Phased roadmap. What happens in the next quarter. What happens in the quarter after that. What happens in the next year. Notice I didn't say "what happens in the next three years." AI capability is shifting too fast for a three-year roadmap to be credible, and honestly, your firm's readiness to absorb it will change materially over eighteen months too. Keep the horizon to twelve months, with the first quarter being the most detailed. This isn't a comprehensive programme plan - it's a sequenced set of commitments that tells the board: here's what we're doing first, here's what follows, and here's why that order makes sense.
6. Investment requirements. What the phased roadmap costs - in money and in people's time. Express these as ranges where precision isn't warranted, which at this stage is almost everywhere. "The first pilot will require £40k-60k of external support and approximately 15-20 days of internal resource over 90 days" is more honest and more useful than a precise figure that implies a certainty you don't have. Boards in professional services firms respond better to honest ranges than to false precision. They've been burned by the latter too many times.
7. Success measures. Three to five specific, measurable outcomes that will tell you whether the strategy is working. "Improve efficiency" is not a success measure. "Reduce average proposal turnaround time from five days to two days within six months of pilot launch" is. Reviewed quarterly. Not annually. I'll come back to why.
This bit might save you the most time. If you're under pressure to produce this document quickly, knowing what you don't have to include is as valuable as knowing what you do.
Vendor-specific technology recommendations. Don't name platforms, tools, or providers in the strategy document. "We recommend implementing OpenAI's GPT-4 for document summarisation" belongs in a procurement process, not a strategy. The moment you bake a specific vendor into the strategy, you've created a document that becomes obsolete when the market shifts - which, in AI, means roughly every four months. Keep the strategy at the level of capability and outcome. Let the procurement decisions follow.
Exhaustive use case lists. Someone on the team has done genuinely good research and identified 30 or 40 potential AI applications. There's a real gravitational pull toward including all of them, because it demonstrates thoroughness. Don't. A strategy that says "here are 40 things we could do" is a strategy that says "we couldn't decide." Stick to three to five. Put the full list in an appendix if you must - but the main document should show the outcome of prioritisation, not the input to it.
Technical architecture. The board audience for this document is not reading it to understand system design. They don't need to know about API layers, vector databases, or model fine-tuning. If the technical architecture matters for the pilot, it goes in the pilot brief. The strategy document should be readable by a managing partner who doesn't know what a large language model actually is - and who, frankly, shouldn't need to in order to make the governance decisions the document asks of them.
Forward projections beyond 18 months. I know some strategy frameworks want you to paint a picture of where the firm will be in 2028. In most domains, that's reasonable. In AI, it's fiction. The capabilities available to you in 18 months will be materially different from today, and so will the regulatory landscape. Plan to 12 months with any specificity. Acknowledge the 18-month horizon. Beyond that, you're writing creative fiction, and the board will know it.
Non-negotiable, in my view. If your AI strategy document is longer than 15 pages, it will not be read in full by the people who need to read it. I've tested this informally with partners and board members across a dozen firms, and the pattern is consistent: they read the executive summary, skim the headings, and maybe - maybe - read the investment section. Everything else gets filed under "I'll come back to it," which is polite code for "I won't."
So. Two-page executive summary, written specifically for the board member who will not read the full document. Then seven sections, each capped at one to two pages. If a section requires more than two pages to explain, that section needs simplifying, not expanding. The opportunity identification section can carry an appendix with deeper detail, but the main document should summarise the decisions, not the deliberation.
Every page you add reduces the probability that the document changes anything. That's a bad trade.
A quick word on this, because I've watched people get it wrong in ways that were entirely avoidable. A board presentation of your AI strategy is not a slide-by-slide walkthrough of the document. It's a 20-minute conversation - genuinely, 20 minutes, not 45 - structured around three questions.
Where are we now? Present the current state as a readiness profile, not a score. The board needs to understand the honest starting position.
What do we want to do first, and why? Walk through the prioritised opportunities with enough context that a non-technical board member understands the commercial rationale. Skip the technical detail.
What do we need from you? This is the bit most presenters fumble. The board exists to make governance decisions, approve investment, and hold people accountable. Tell them specifically what you need them to decide, what you need them to fund, and who you need them to hold responsible. Leave a board meeting without those three things resolved and you've had a status update, not a decision.
And one thing I've learned the hard way: the conversations that matter happen before the board meeting, not during it. If you're presenting an AI strategy cold to a board that hasn't been warmed up through individual conversations, you're taking a risk you don't need to take. Nobody likes surprises in a boardroom.
This is where most AI strategy documents actually fail. Not in the writing, but in what happens after.
A strategy document reviewed once a year will be out of date within three months and ignored within six. AI moves too fast, your firm's understanding evolves too quickly, and the competitive landscape shifts underneath you while the document sits unchanged on the shared drive.
The fix is a quarterly review cadence. At Distinction, we use a framework called WHNN® - the What and the How, for the Now and the Next - that runs on a repeating quarterly cycle. I've seen it work well with professional services firms specifically because it's built around the kind of structured, recurring governance conversations those firms are already comfortable with. The "Now" dimension is particularly relevant here: what was committed to last quarter? What was actually delivered? What's changed externally - in AI capability, in the regulatory environment, in what competitors are doing? And based on all of that, what are the adjusted priorities for the next quarter?
The key principle: the quarterly review should produce a one-page update to the strategy document, not a new strategy. Over four quarters, those cumulative updates tell a story of a firm that is learning, adapting, and making progress - which is what the board actually wants to see. And counterintuitively, it reduces governance overhead over time, because you're making small adjustments each quarter rather than letting the strategy drift until someone commissions a full rewrite.
Think of it like maintaining a house. A bit of upkeep each season is cheaper - and less stressful - than waiting until the roof leaks.
Look, I know the reality. Most AI strategy documents are produced because someone senior asked for one, and the team responsible just wants to get it done and move on. That's human nature, and I'm not going to pretend otherwise.
The firms I've seen make genuine progress with AI - the ones where the strategy actually translated into working tools that saved people time, won better work, or improved the client experience - treated the document as an operating instrument, not a deliverable. Short enough to be read. Specific enough to be acted on. Structured so the quarterly rhythm kept it honest.
If you want a template for the AI strategy document that covers all seven sections, applies the 10-15 page constraint, and includes the quarterly review format - ready to populate for your firm's specific situation - [download it here].
And if you'd rather produce the strategy document through a facilitated session that structures the current state assessment and opportunity prioritisation in a single day - [book an AI strategy session]. We've run these for professional services firms from 100 to 500 people, and the output is a document your board will actually read. Which, as we've established, is the whole point.