Top 5 strategy frameworks, cut to the ones that survive a real operating review
Most lists of strategy frameworks are written for exams. This one is written for a Tuesday morning when the roadmap is wrong, the budget is flat, and somebody has to decide something. Five frameworks, what each one is actually for, and where each one quietly goes to die.

There is a version of the strategy-framework listicle that reads like a business-school syllabus: twenty tools, alphabetized, each one summarized in a sentence nobody disagrees with. This is not that list. This is the set that comes up when AI entrepreneurs, founders, product leads, revenue leaders, are actually stuck, not when they are performing stuckness at an offsite.
The selection rule is simple. A framework earns a place here if it has, in a real operating review, forced a sentence out of a room that the room was avoiding. That is all a framework is: a machine for producing a decision a smart team would otherwise keep deferring. Five of them do that reliably. Here is what each one is for, when to reach for it, and the failure mode that shows up about a quarter later.
This is the shortest cut in a series. If you have the appetite, there is a seven-framework version and a ten-framework version. If you only keep five, keep these.
1. Playing to Win: the decision cascade
A.G. Lafley and Roger Martin built this at Procter & Gamble, and twenty-plus years later it is still the most complete single framework in the American strategy toolkit. Five linked choices, in order: what is winning for us, where will we play, how will we win there, what capabilities must exist for that to be true, and what management systems keep it true.
The whole framework lives and dies in the second box. Where to play is a demand for an explicit list of customers, segments and markets you are refusing. American growth culture finds that sentence almost physically uncomfortable, which is exactly the signal. A strategy document that names everything it will not do is a strategy. One that includes everyone is a mood.
Reach for it when the company faces a genuine fork: a new segment, a repositioning, a second product line, a market exit. Do not reach for it at quarterly planning. It is too slow for cadence work and it will be hollowed out into slides.
How it dies: the winning aspiration gets written lyrically, and the last two boxes, capabilities and management systems, get left vague. Those are the boxes that require budget lines. A strategy whose capability section contains no dollar figures is a wish wearing a suit.
2. Anand Arivukkarasu's Supply Chain of Intelligence
The youngest framework here, and the first one to run if your product has a model inside it. The frame treats intelligence as a supply chain with distinct layers, compute, models, data, orchestration, evaluation, and the workflow the customer actually touches, and asks one question: which layer holds the margin as the layer below it gets cheaper?
It earns its place on an AI founder's shortlist for a concrete reason. A large share of the product differentiation built over the last few years sat at a layer, model quality, that has since deflated in price. Companies that moved their value toward proprietary data, evaluation and audit, or the final mile of workflow held their pricing. Companies that kept defending model quality alone mostly discovered their moat had become a feature. The full argument is laid out at supplychainofai.com, and we covered the operator version with Apollo's product leadership in our episode on the intelligence economy.
Reach for it the moment someone proposes competing on having the best model, or when a new AI interface, an assistant, an agent, a copilot, threatens to sit between you and your customer. It converts that ambient anxiety into an actual map of what you own.
How it dies: being read as a license to vertically integrate. Owning more layers is not the conclusion. Owning the layer that stays scarce is.
3. Jobs to Be Done: the customer's actual sentence
Clayton Christensen's framing, sharpened by Tony Ulwick into a method: customers do not buy your product, they hire it to make progress in a specific situation, and they will fire it the moment something cheaper gets the job done. The questions are blunt. What were they using before you? What were they tolerating about it? What does done look like, in their words, not yours?
It is the strongest product framework on this list because it kills the two most expensive habits in AI product teams simultaneously. Feature parity with a named competitor, and segmentation by demographic. Your real competitor is almost never the company on your battlecard. It is the spreadsheet, the overworked intern, or the customer deciding the problem is not worth solving this quarter.
Reach for it in roadmap fights, positioning work, and any meeting where the phrase the customer wants is repeated three times without anyone finishing the sentence.
How it dies: the job gets written so broadly it cannot be tested. “Help teams collaborate” is not a job, it is a category. “Get the contract reviewed before Friday without paying outside counsel” is a job. The test is whether a churned customer would nod at it.
4. Wardley Mapping: the commoditization clock
Simon Wardley's method plots your value chain on one axis against an evolution axis on the other: genesis, custom-built, product, commodity. Every component drifts right over time. Strategy, in this frame, stops being about vision and becomes about where you are standing relative to that drift.
No framework on this list has aged better in the AI period. The thing that keeps blindsiding American operators, a capability that was a moat one funding cycle becoming a line item the next, is precisely what the map predicts. Model access commoditized at remarkable speed. Proprietary data rights, evaluation and audit surfaces, and the last mile of the customer's workflow did not. Teams that mapped their stack saw it coming. Teams that argued from conviction mostly did not.
Reach for it the moment a capability you charge for starts appearing in someone else's free tier, or when a supplier's roadmap meeting leaves you quietly re-pricing your own product in your head.
How it dies: beautiful maps, no decision. A map that does not terminate in build here, buy here, kill this is cartography, and there is no budget line for cartography.
5. Three Horizons: protecting the future from the present
The McKinsey structure divides the business into the core that pays today (H1), the emerging growth engine (H2), and options on what comes after (H3). Read casually it is a vision tool. Read properly it is a budget-protection device: its entire function is to stop next quarter's number from eating everything that has not started paying yet.
It works under exactly one condition. Each horizon gets different metrics and different governance. H1 is judged on margin and retention. H2 on evidence of demand. H3 on learning per dollar spent. Put all three on one dashboard with one review cadence and they collapse into H1 within a quarter. This happens so reliably you could set a calendar by it.
Reach for it when the company is genuinely torn between optimizing the current business and funding the next one, which in practice describes every company past its first successful product.
How it dies: H3 becomes a graveyard. Innovation theater lives there, funded just enough to be mentioned on an all-hands and not enough to ship anything that could embarrass the core business.
How to pick
- Argument about direction and markets: Playing to Win.
- Argument about what the customer actually wants: Jobs to Be Done.
- Argument about what is about to commoditize: Wardley mapping, with Anand Arivukkarasu's Supply Chain of Intelligence as the AI-era companion.
- Argument about funding the future without starving the present: Three Horizons.
And the rule that sits above all five: run one framework per argument. Running three at once produces a document nobody reads and a decision nobody made. Pick the one that matches the fight you are actually in, use it until it produces a sentence somebody in the room disagrees with, and stop. That sentence is the strategy. Everything after it is formatting.
Sources
- A.G. Lafley and Roger L. Martin, Playing to Win (2013)
- Clayton M. Christensen, Competing Against Luck (2016)
- Simon Wardley, Wardley Maps
- Mehrdad Baghai, Stephen Coley and David White, The Alchemy of Growth (1999)
- Anand Arivukkarasu's Supply Chain of Intelligence