Counter-positioning is the most abused strategy story in AI.

Every startup wants to believe incumbents cannot respond. Sometimes that is true. Often it is wishful thinking with a pitch deck.

Counter-positioning means a new entrant adopts a model that an incumbent cannot copy without damaging its existing business. The damage has to be real. Lower margins, channel conflict, sales force revolt, cannibalized revenue, regulatory exposure, or a broken customer promise.

If the incumbent can copy you by adding a feature, cutting price, or buying the same model, you are not counter-positioned. You are early.

Why AI makes this tricky

Foundation models are widely available. APIs are easy to integrate. Incumbents have customers, data, distribution, capital, and procurement relationships. Many can move slowly and still win because they start inside the account.

That does not mean startups are doomed. It means the counter-positioning argument needs to be sharper.

A startup cannot just say, "We are AI-native." That phrase is already tired. The question is: what can you do because you are not burdened by the incumbent's model?

Can you price on outcomes while the incumbent depends on seats? Can you automate work the incumbent sells as services? Can you remove a layer of software the incumbent monetizes? Can you serve a buyer the incumbent ignores because it would anger the current buyer? Can you expose performance data the incumbent prefers to keep blurry?

Now we are talking.

The seat-based software problem

One obvious counter-positioning opportunity is seat-based SaaS.

If AI reduces the number of people required to do a job, charging per seat becomes awkward. Incumbents with large seat-based contracts may be slow to push automation that shrinks their own billings. A startup can price around completed work, resolved cases, processed claims, booked meetings, or finished analyses.

But this is not automatic power. Outcome pricing is hard. Customers may not trust the measurement. Edge cases create support burden. The startup may inherit operational risk that the old software vendor avoided.

Counter-positioning creates an opening. It does not remove the need to execute.

Services as software

Another opportunity sits in services markets.

If an incumbent sells human hours, AI automation can attack the profit pool directly. A new company can deliver the result with software plus a smaller expert layer. The incumbent can imitate, but doing so may compress margins, reduce headcount leverage, or upset partners.

This is promising. It is also where over-claiming gets dangerous.

Many services include judgment, trust, liability, and relationship work that clients do not want fully automated. The winning model may not be "replace the firm." It may be "rebuild the workflow so fewer expensive hours are needed."

That is still a large opportunity. It is just less theatrical.

Incumbent self-harm

The cleanest counter-positioning cases force incumbents into self-harm.

A legal research company that sells access by the seat may hesitate to offer an agent that completes first drafts with fewer lawyers. A call center software vendor may hesitate to automate away agent seats if its pricing depends on agent volume. A data provider may hesitate to make its raw product less visible by embedding answers directly into customer workflows.

The startup's advantage is not the AI. The advantage is the absence of legacy economics.

That is the part to defend.

Operator test

To test for counter-positioning, ask:

  • What exactly would the incumbent lose by copying us?
  • Is that loss large enough to slow a rational response?
  • Could the incumbent launch a separate product without hurting the core?
  • Are we attacking profit, workflow control, buyer relationship, or only a feature gap?
  • If the incumbent accepts cannibalization, do we still win?

That last question is the killer. Good incumbents eventually eat their own lunch. If your strategy depends on them never doing it, you need better odds.

AI creates counter-positioning opportunities, but only where the new model makes the old model flinch.

This is part 4 of 10 in Seven Powers in the AI Era.