Most people who sit down to build a strategy with AI make the same mistake in the first thirty seconds: they ask it to write the strategy.
You’ll get something back. Well-structured, confidently worded, and completely generic — a plan that would fit any firm in your industry, which means it fits none.
The alternative is to use AI the way you’d use a good facilitator: let it run the process, ask the questions and hold the standard, while you do the deciding. An afternoon’s work, and you get something you’ll still defend in June.
Here’s what that session looks like, phase by phase, with the real questions and the quality bar each phase holds. It’s what the OGSM Strategy Builder does — but the process is the process, and you can get value from this post without buying anything.
What do you need before you start?
Not much, but the numbers matter. Have these to hand:
- Revenue — last full year and the one before, so growth is visible.
- Margin — gross, and operating if you have it.
- Growth rate — actual, not aspirational.
- Customer concentration — what share of revenue your largest customer represents, and your top five.
- The one operational number your business runs on. Utilisation for a services firm, lead time for a manufacturer, covers per night for a restaurant, renewal rate for anything on contract. You know yours.
Then two more things: two to four hours of real work, which almost nobody does in one sitting, and a working answer to why now? Nobody writes a strategic plan because it’s January. Something changed, or is about to, and that’s the real subject of the session.
If a number isn’t to hand, estimate it, say so, and move on. It gets flagged later as something to verify.
My worked example throughout: Marlow Instrumentation Services, a fourteen-person Midlands firm calibrating lab instruments for pharmaceutical and food-testing customers. Revenue £2.4m, growing 3% a year, gross margin sliding from 41% to 36%, largest customer 22% of revenue.
Phase 0: What does the diagnosis actually ask you?
Ten to twenty minutes, and the phase everyone wants to skip.
It opens with real questions — what you do, who buys from you, how you’ve changed. Then the one that matters: what made you want to do this now?
Then scope (whole business or one function?) and horizon — a 3–5 year strategic plan or a 1-year operating plan. Pick one. It governs how ambitious the objective can be, whether timelines read 2028 or Q3, and whether reviews are quarterly or monthly.
Then the Five-Answer Test:
Ask five of your leaders, separately, so they can’t hear each other: what is our strategy?
You’ll get five answers. Each reasonable, each confident, and not one quite the same. Solo? The equivalent is: could your best customer say what you’re trying to be?
Then the Seven Cardinal Sins self-diagnosis, asked one at a time. Are you deciding from a picture of your market that’s a few years old (Ignorance)? Do five leaders give five answers (Discord)? Could someone on the floor say what the company is trying to do (Confusion)? Does the daily work quietly contradict the stated strategy (Incongruity)? Is there someone who has to carry this who doesn’t believe in it (Doubt)? Does the plan need capabilities you don’t have yet (Incompetence)? Is it agreed, pinned up, and not moving (Inertia)?
Most businesses recognise three or four. That’s normal, and worth saying out loud, because people get defensive here. The ones you name become the watchlist for every phase that follows.
Finally: list everything the business is currently working on. Everything taking real time or money. It’ll be longer than you expect — twenty items is common. Nothing gets resolved now. That list comes back in Phase 5, and it’s the argument.
Phase 1: How does the reality check work?
Key figures first — the numbers you prepared. Then a SWOT, built by conversation rather than form-filling.
The questions have teeth. Name a strength and you’ll be asked who else in your market would claim the same thing? Give a number and you’ll be asked where it came from. Each quadrant caps at three to five items: if a strength doesn’t make your top five, it isn’t a key strength.
Marlow’s SWOT surfaced the pairing that mattered — their biggest weakness (a price list that hadn’t kept up with wage inflation) sitting against their biggest threat (two national providers undercutting them per instrument). That pairing became a strategy three phases later.
Where your situation warrants it you’ll be offered PESTEL, market trends, Porter’s Generic Strategies, Five Forces or the 5P marketing mix. Where it doesn’t, you won’t. A fourteen-person firm doesn’t need Five Forces to work out that price competition has arrived.
Phase 2: How do you land on an objective you’d defend?
Twenty to forty minutes. It’ll feel long. It should — everything downstream is derived from this one sentence.
You’ll be pointed back to your why now? answer, then asked a present-tense question that people find easier than the future-tense one:
When a customer chooses you over the alternative today, why do they?
The honest answer is usually the raw material for the second half of your objective.
Then three options — three genuinely different bets, not three rewordings — each with what it commits you to and what it rules out. You choose, or say the real one is none of them.
Marlow’s first attempt was typical:
❌ “Grow the business by delivering excellent service to our clients.”
Three problems. “Grow” is an outcome, not a direction. “Excellent service” is a claim every competitor makes. And swap in a competitor’s name — it still reads fine, which means nothing was chosen.
After three drafts:
✅ “Become the default calibration partner for regulated pharmaceutical and food-testing labs in the Midlands by making audit-ready documentation, not price per instrument, the reason customers stay.”
That names a segment, names a mechanism, and rules something out — price-led work, about a third of their current enquiry flow.
Then five quick tests: substitute a competitor’s name (does it still work?); could five people repeat it the same way after hearing it twice?; what does it rule out?; what would someone do differently on Monday?; and is it ambitious enough to be worth it, grounded enough that nobody laughs? More on the craft in how to write a great objective.
Phase 3: How do you turn that into goals?
Twenty to thirty minutes. Three to five SMART goals, each written so it can be checked without a conversation: [metric] from [baseline] to [target] by [date].
It starts with the twelve-month question — we meet again in a year, it’s gone well, what’s different? — then two relentless follow-ups: how much? and is that ambitious enough? The second is where the real number appears. People’s first answer is almost always their safe answer.
❌ “Improve customer retention.”
✅ “Contract renewal rate from 81% to 92% by 31 December.”
Two rules shape the set. At least one financial goal — every strategy eventually shows up in the money. And goals covering both halves of your objective, the “what” and the “how”. Marlow’s renewal and revenue goals cover the what; regulated-lab revenue share and documentation turnaround cover the how. A plan with only “what” goals can be hit by accident, doing all the things you said you’d stop.
Then the gate, which is the most useful question in the method:
If we achieve exactly these goals — and nothing else — have we succeeded?
If something is missing, write a goal for it. If a goal could be missed entirely and you’d still say the objective was met, it’s a metric you like, not a goal — it belongs in the Dashboard. See Goals vs Measures.
Phase 4: What makes the strategies phase uncomfortable?
Thirty to fifty minutes, and the phase that matters most. Everything before this could be done by a thoughtful person with a spreadsheet. This is where something gets given up.
You’ll be pushed back to the SWOT pairings — your biggest weakness is X and your biggest threat is Y; what specifically are you going to do about that? You’ll generate six or eight candidates and then be forced to cut to three to five. The cut is the work.
Each one has to pass the tactic test. A strategy sets a direction; it does not spell out the work. If it still leaves a real “but how, exactly?” for the next phase to answer, it is pitched at the right level. If it already specifies the work in full, it is an initiative and belongs in Phase 5.
❌ “Improve our marketing.” — Too vague. Costs nothing to agree to.
❌ “Redesign the website.” — Too specific. That’s an initiative.
✅ “Win regulated-lab accounts from the national providers by selling audit-readiness and documentation turnaround instead of price per instrument.” — Owner: Priya.
Every strategy gets a named owner. Not “the team”. A person who has to answer for it at the review.
Then the We WILL / We will NOT list, which takes ten minutes and turns intentions into a decision:
| We WILL | We will NOT |
|---|---|
| Focus on regulated pharma and food-testing labs | Quote on single-instrument, price-only tenders |
| Compete on documentation and turnaround | Compete on price per instrument |
| Build accreditation depth in two instrument classes | Cover every instrument a customer owns |
The right-hand column is the one that matters and the one that gets skipped. When you resist, you’ll get prompts like what would your competitor be pleased to hear you’d decided to keep doing?
For Marlow, that column meant walking away from roughly £180k of low-margin work. Someone had built that revenue. That’s what a real trade-off feels like.
Phase 5: How do the measures get built?
Thirty to forty-five minutes. Per strategy, two structurally different things, kept explicitly apart.
The Dashboard — 2–4 KPIs, each with a definition, baseline, target, frequency and data owner. At least one leading, one lagging. Six to twenty across the plan.
The Action Plan — 2–3 initiatives that cause the strategy to work, each with a caretaker and dates.
The speedometer doesn’t make the car go faster. Without the Dashboard you’re flying blind; without the Action Plan you have targets and no engine. Full method in the complete guide to OGSM Measures.
Three questions build a dashboard. If this strategy is working, what would change? Then the “so what?” filter — if this number moves, what decision follows? If none, drop it. Then: what would you need to see in ninety days, before the annual numbers land? That last one finds your leading indicator.
❌ “Number of sales calls made.” — Activity. Tells you what your team did, not whether it worked.
✅ Lagging: revenue from regulated labs — £912k to £1.44m — monthly — Priya.
✅ Leading: quotes issued to regulated labs — 4/month to 12/month — monthly — Priya.
✅ Initiative: rebuild the certificate pack so it’s audit-ready without customer follow-up — Dan — January to end of Q2.
Anything without a baseline gets refused. If the baseline genuinely doesn’t exist, establishing it becomes the first initiative under that strategy.
Then the capacity conversation, the antidote to Inertia. Your Phase 0 workload list comes back and every item goes into one of three buckets: continues, paused, stopped.
You’ve got nine new initiatives on top of eighteen existing ones. Which existing ones stop? Not “deprioritised” — stopped. If everything stays, we execute nothing.
Focus is not about adding. It’s about removing.
Phase 6: What does the stress test look for?
Fifteen to twenty-five minutes, delivered in three parts: what holds, what I’d challenge, and what I can’t verify.
Mechanical checks first — counts, missing baselines, strategies without owners, activity metrics wearing outcome metrics’ clothing, an empty “will not” list. Then judgement: the competitor test, the arithmetic test (do these strategies plausibly deliver these goals, or is it 28% growth against three efficiency initiatives?), the resourcing test, the capability check and the belief check.
That third section is usually missing from AI strategy work, and it’s the honest bit. It has read everything and lived none of it. Your market size, a competitor’s intent, whether your team can absorb nine initiatives — those come back to you to check before the plan goes live.
Where you disagree with a finding, it’s recorded as an accepted risk with your rationale and a watch-for. Not a defeat — a decision made with eyes open.
Phase 7: How does it get delivered and cascaded?
Fifteen to thirty minutes. The one-pager is built in the formats you want — HTML to share, Excel to track, PowerPoint to present, PDF to pin up.
Then cascade guidance, if you have layers: a higher-level Measure becomes a lower-level Goal, cascaded to business and function level only, never to individuals. The named failure is cascading by copying rather than translating — a divisional plan that restates the corporate strategies in different words hasn’t cascaded, it’s photocopied.
Then the communication plan — to whom, what, how, when, by whom — with three to five key messages you could repeat from memory. And the review calendar: monthly for a one-year plan, quarterly for a three-to-five-year one, with every meeting in the diary before the session ends. Agendas in how to run an OGSM review meeting.
It closes by asking for three concrete things: the first review date, the first thing that stops, and the first visible win.
Phase 8: What happens at the review?
This phase repeats, and it is the one that decides whether the other eight were worth the afternoon.
Before each review you bring the workbook back. It goes through the plan strategy by strategy — not metric by metric, because you think in strategies and a list of thirty numbers produces thirty guesses. Where are the numbers, and where are the initiatives?
Then it does three things you would probably skip on your own.
It compares against the plan rather than against last month. Not “did it go up?” but “is it on the path we said?” If the goal is a 30% improvement over twelve months and four months in you have moved 3%, it will tell you that you are a third of the way through the time and a tenth of the way through the distance. That sentence is more useful than any adjective.
It checks the “we will not” list. Strategies are rarely reversed in a meeting. They get reversed one exception at a time. “You said you’d stop taking price-led work. Two of your last four wins were price-led. Either the strategy has changed or the behaviour needs to.”
And it flags the item that has been amber for four months and never discussed — because amber feels survivable, and an initiative holding the same non-green status for three cycles is not a status, it’s a decision nobody is making.
You get a prepared agenda with at-risk items first, and the short list of things that need a decision rather than a status update. Afterwards, you come back and say what was decided, and it goes into the workbook.
One rule it will hold you to: don’t change a strategy because it’s uncomfortable. Give it two or three review cycles. Most strategies look wrong at month three, because the cost has arrived and the benefit hasn’t. Folding then is the commonest way a sound plan dies. Change when the ground has genuinely moved — a competitor arrives, a customer concentration risk becomes an event — not when you have simply gone off it.
How does saving and resuming work?
At the end of every phase your progress is written to a workbook file and handed to you. Save it. To come back — next evening, next month — start a session, hand the file back, and you’ll be told in two sentences where you left off. Nothing you’ve decided gets redone.
Use it. A plan built across four evenings beats one rushed in an afternoon, and Phase 4 benefits from sleeping on it.
Frequently asked questions
How long does it actually take?
Two to four hours of real work — nearer two if your numbers are to hand and you decide quickly, nearer four if you take the analysis seriously. Most people split it across two or three sittings, which is what the save-and-resume is for. Phases 2 and 4 take the longest and should. If you finish in twenty minutes, you filled in a template.
Can I use this if I already have a strategy?
Yes — often the better use. Run Phase 0 and Phase 6 against what you have, then re-enter wherever your plan first breaks down. Usually Phase 4, because most existing plans have goals and initiatives but no actual choices.
Do I need my leadership team in the room?
Not to build the draft. But be clear what you’re producing. If you own the business, this is a decision. If you’re a manager in a company of four hundred, it’s a strong draft to take into a real conversation.
What if I don’t have baselines for everything?
Write [baseline: to be established by DATE] rather than leaving it blank, and make establishing it the first initiative. A target without a baseline is unfalsifiable — in nine months nobody will agree whether it was hit.
Will the AI just agree with everything I say?
It shouldn’t, and a good process is designed against it. Expect to be told an answer could apply to any company in your industry, and to be asked what you’d stop. If the session felt comfortable throughout, something was avoided.
Can it decide my strategy for me?
No, and you shouldn’t want it to. It doesn’t know your market, your customers, or what happened last time you tried this. It drafts options; choosing is yours. A strategy you didn’t choose is one you won’t defend when it costs you something.
What happens after the plan is built?
Phase 8, every month or quarter, for as long as the plan is live. Twenty to thirty minutes with the workbook before the meeting you run with your team. See how to run an OGSM strategy review meeting for the agenda itself.
Is this different from prompting Claude directly?
Yes, in the way that matters: consistency of standard. A prompt gets one good answer. A process holds the same bar across eight phases, remembers Phase 0 when you contradict it in Phase 5, and refuses the shortcuts. If you’d rather work from prompts, how to use AI to build your OGSM has them.
Where to go from here
Nothing here requires a purchase. Open Claude, work the phases in this order, hold yourself to the quality bars, and you’ll produce something far better than a filled-in template — because the hard part is the questions, not the format.
Read the common OGSM mistakes alongside it, and 30 OGSM examples across six industries if you want to see finished ones first.
And if you’d rather not have to remember the questions or hold the standard yourself, the OGSM Strategy Builder does all eight phases with you and is available in the shop.
Either way, do it before January. The plan you write in a quiet week is worth three you write under pressure.
Rock on.