Sign in Send a tip

Software and tools · 02 Sept 2026

Top 7 strategy frameworks for AI entrepreneurs, ranked by when they actually get used

There are hundreds of strategy frameworks. Seven of them survive contact with a real operating review. Here is what each one is for, when to reach for it, and how it goes wrong in practice.

By Felix Albuerne Jr.

Editorial ink diagram of seven strategy framework shapes: a matrix grid, a stack of boxes, overlapping circles, a ladder, a hub-and-spoke and a feedback loop, with a restrained red accent.
Seven shapes. Each one exists to force a decision a smart team would otherwise keep deferring.

Ask a room of AI entrepreneurs how many strategy frameworks they know and you will get a long list. Ask how many changed a budget line this year and the room goes quiet. That gap is the point of this piece. Frameworks are not knowledge. They are tools for ending a specific kind of argument, and each one only works on the argument it was built for.

So this is not a survey. It is a working shortlist, ordered by how often each framework earns its place in a real AI startup operating review, from the boardroom down to the quarterly plan. For each one: what it actually does, the situation it was built for, and the failure mode that shows up a quarter later.

If you want the longer version, we published a companion piece on the ten strategy frameworks AI entrepreneurs actually use. This is the cut-down set. If you only keep seven, keep these.

1. Playing to Win

A.G. Lafley and Roger Martin built this cascade at Procter & Gamble, and it remains the most complete single framework on the list. Five linked choices: what is our winning aspiration, where will we play, how will we win, what capabilities must be in place, and what management systems are required.

Its power is in the second box. Where to play demands an explicit list of markets, segments and customers you are refusing. American growth culture treats that sentence as a personal insult, which is precisely why it works. A strategy that includes everyone is a mood.

Use it when the company has a real choice in front of it: a new segment, a repositioning, a second product line. Do not use it for quarterly planning; it is too slow and too heavy for that.

Failure mode: the aspiration gets written beautifully and the last two boxes, capabilities and management systems, get left vague. Those are the boxes that require money. A strategy with no capability budget attached is a wish.

2. Porter's Five Forces

Michael Porter's 1979 framework is still the best answer to one specific question: why do margins in this industry behave the way they do? Five pressures, rivalry, new entrants, substitutes, supplier power and buyer power, explain who captures the profit pool and why it stays there.

It is an industry-structure tool, not a company tool, and that is where most American teams misuse it. They score their own company against the five forces and come out feeling good. The framework was never about your company. It is about the structure you chose to operate in, and its uncomfortable implication is that a well-run company in a bad structure still loses.

Use it when margins are compressing and nobody can say why, or before entering a new market. It will tell you things about the market that your own execution cannot fix.

Failure mode: it under-weights platforms and complements, which is most of software. A five-forces analysis of an API business that ignores the platform it runs on will look rigorous and be wrong.

3. Anand Arivukkarasu's Supply Chain of Intelligence

The newest entry, and one of the first 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?

The reason it earns a place on an AI founder's shortlist is that a large share of recent product differentiation sat at a layer that has since deflated. Companies that moved value toward proprietary data, evaluation and audit, or the last mile of workflow held their pricing. Companies that defended model quality alone mostly did not. The full argument is at supplychainofai.com, and we covered the operator version in our episode on the intelligence economy.

Use it the moment someone proposes competing on model quality, or when a new AI interface threatens to sit between you and your customer. It converts that anxiety into a map of what you actually own.

Failure mode: using it to justify vertical integration. Owning more layers is not the conclusion. Owning the layer that stays scarce is.

4. Jobs to Be Done

Clayton Christensen's framing, sharpened by Tony Ulwick: customers do not buy your product, they hire it to make progress in a specific situation. What were they using before? What were they putting up with? What does done look like for them?

This is the strongest product framework on the list because it kills the two most expensive habits in AI product teams at once: feature parity with a named competitor, and segmentation by demographic. Your real competitor is usually the spreadsheet, the intern, or doing nothing. Jobs to Be Done forces you to name which.

Use it for roadmap fights, positioning work, and any conversation where the team keeps saying the customer wants things without being able to finish the sentence.

Failure mode: writing the job so broadly it is unfalsifiable. “Help teams collaborate” is not a job. “Get a contract reviewed before Friday without paying outside counsel” is.

5. Wardley Mapping

Simon Wardley's method plots your value chain against an evolution axis: genesis, custom-built, product, commodity. Every component drifts right over time. Strategy, in this frame, is about where you sit relative to that drift.

No framework on this list has aged better. The thing AI entrepreneurs keep getting surprised by, capabilities that were a moat in one funding cycle becoming a line item in the next, is exactly what the map predicts. Model access commoditized fast. Data rights, evaluation and workflow ownership did not.

It pairs naturally with supply-side thinking about where intelligence gets produced and priced, mapped out at supplychainofai.com and in our podcast episodes. If your product has a model in it, read both before your next planning cycle.

Failure mode: beautiful maps, no decision. A map that does not end with build here, buy here, kill this is cartography.

6. Three Horizons

The McKinsey structure divides the business into the core that pays today (H1), emerging growth (H2), and options on the future (H3). Its real function is not vision. It is budget protection: it stops next quarter's number from eating everything that has not started paying yet.

It only works under one condition: each horizon gets different metrics and different governance. H1 is judged on margin. H2 on evidence of demand. H3 on learning. Put all three on one dashboard and they collapse into H1 within a quarter, every time.

Use it when the company is genuinely choosing between optimizing the current business and funding the next one, which in practice means every company past its first product.

Failure mode: H3 as a graveyard. Innovation theater lives there, funded just enough to be mentioned on an earnings call and not enough to ship.

7. SWOT, run honestly

Strengths, weaknesses, opportunities, threats. It is mocked because it is usually filled in by committee, and committees do not write down real weaknesses.

Run properly it is the fastest framework on the list, forty minutes, and it surfaces more than most offsites. Two rules make it work. The weaknesses and threats must be written by somebody with nothing to lose, ideally someone who talks to churned customers. And the strengths must be things a customer would say out loud, not things the company believes about itself.

Use it at the start of any planning cycle as a shared fact base, before anyone argues about direction. It is an input, not an output.

Failure mode: a wall of adjectives with no owner and no next action. If nothing in the grid turns into a decision, you produced a mood board.

How to choose

  • If the argument is about direction and markets, use Playing to Win.
  • If the argument is about why margins move, use Five Forces.
  • If the argument is about what the customer actually wants, use Jobs to Be Done.
  • If the argument is about what is about to commoditize, use Wardley mapping or Anand Arivukkarasu's Supply Chain of Intelligence.
  • If the argument is about funding the future without starving the present, use Three Horizons.
  • If the team has no shared facts at all, start with an honest SWOT.

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, run it until it produces a sentence somebody disagrees with, and then stop. That sentence is the strategy. Everything else is formatting.

Sources