← Back to the redesign toolkit

Appendix B · Chapter 7 · AI That Rebuilds the Business

The Moat-vs-Parity Test

This is the complete test from Chapter 7, in usable form — it tells you whether what you built is a moat (an edge a competitor can’t cheaply copy) or just parity (a real improvement that anyone with the same tools can match). Run it on a specific redesign candidate — not on your ambitions in general — after the Five-Question Screen and before serious money moves. Then re-run it at every review cycle, because a yes today is not a yes at the next review.

The standing instruction. Score this with your Distance source in the room — the one person with no incentive to grade generously. The failure mode of this test is the grader: everyone scores their own strategy as the rare exception, for the same reason 93% of drivers rate themselves above the median (Svenson, 1981). Self-administered, this test returns “moat” almost every time. That result is worthless.

Part one — the VRIN test (Barney, 1991)

Four gates, in order. A no at any gate ends the advantage claim.

V — Valuable

Does the redesigned process measurably reduce cost, increase willingness to pay, or open revenue you couldn’t reach before? Not “improves efficiency” — a number, against your baseline.

R — Rare

Do fewer than a handful of competitors have this? Apply the shared-model rule, and apply it coldly: would a competitor using the same frontier model — the same leading, off-the-shelf AI system anyone can rent — get substantially the same output from substantially the same prompt? If yes, score Rare at zero. No partial credit. Anything a shared tool produced on request is, by definition, available to everyone else on request.

I — costly to Imitate

If a competitor watched your results for a year, what would replication actually cost them? If the honest answer is “a subscription and a consulting engagement,” score it low. Cost to imitate comes from things outside the model — tacit knowledge, exclusive data relationships, an activity system that has to be copied whole.

N — Non-substitutable

Can a competitor get the same customer outcome by a different route entirely? A moat that can be bypassed doesn’t need to be crossed.

Part two — the three core-competence tests (Prahalad & Hamel, 1990)

Test 1 — Market access

Does this capability open more than one market or service line, or is it welded to a single offering?

Test 2 — Perceived customer benefit

Does it make a significant contribution to benefits the customer perceives? Answer from validation data — a real pilot with real customers moving a perception or loyalty metric (Appendix C, Stage 3) — not from internal enthusiasm. An operational win customers never notice, or notice negatively, fails this test. Walmart’s shelf-scanning robots were operationally real; shoppers didn’t want robots in the aisles.

Test 3 — Imitability

Would duplicating this require duplicating your pattern of internal coordination and learning — or just your tooling?

Part three — Porter’s order of fit (1996)

First-order fit

A single improved activity. Easiest to copy. Score any first-order-only redesign as fragile regardless of its VRIN results.

Second-order fit

Activities that reinforce each other — the AI-assisted delivery feeds the data layer that feeds the pricing model.

Third-order fit

Whole-system optimization, where copying one piece without the rest actively hurts the copier. This is why Continental Lite’s partial copy of Southwest made things worse, not better. Hardest to copy; the only order of fit that reliably survives observation.

The verdict

Two honest outcomes:

Moat: passes VRIN including the shared-model rule, contributes a customer-perceived benefit you can evidence, and shows at least second-order fit.

Expensive parity: real spend, real operational improvement, zero durable advantage — because a competitor with the same frontier model and a normal budget replicates it. Most redesigns score here. Parity is not failure; it may be necessary spend just to stay in the game. But it must never be called a moat, because a moat justifies redesign-scale commitment and parity cannot.

How to use it: Run it on one specific redesign candidate, with your Distance source in the room — after the Five-Question Screen, before serious money moves, and again at every review cycle. Print this page →

Want a second set of eyes on this? Chapter 13's offer stands: thirty minutes, nothing for sale, worth the half hour either way.

Book a 30-Minute Call →

The full argument behind this instrument is in AI That Rebuilds the Business.