Back to library

AI Moat Analysis

The base model layer is commoditising fast. Any competitor can call the same API you do. The question for every AI PM is: where is the defensible value? This skill identifies the moat types available to your specific product and designs the investments that widen each one.

Context

The six AI moat types:

Moat typeWhat it isHow hard to replicate
Data moatProprietary training or feedback dataHigh — takes years
Workflow moatDeep embedding in user's daily workflowHigh — requires long-term adoption
Network moatMore users = better product for everyoneVery high — requires scale
Integration moatConnections to user's existing toolsMedium — requires enterprise cycles
Expertise moatProprietary domain knowledge in prompts/fine-tunesMedium — can be replicated
Brand/trust moatUsers trust this product with sensitive dataMedium — slow to build, fast to destroy

Step 1 — Define the product context

Assess: what the product does, who the users are, what data is generated through use, how deeply embedded in workflow, switching cost, top 3 competitors.

Step 2 — Score moat types (0–3)

Score current position and 12-month potential for each moat type. Total score out of 18.

Step 3 — Deep-dive on top two moats

Data moat: What proprietary data do you generate? Why can't a competitor replicate it? What investment widens it? Workflow moat: Where in daily workflow does the product live? What context only your product can see? What is the switching cost? Network moat: Does more usage by one user improve the product for others? What threshold activates the network effect?

Step 4 — Assess competitor moats

Score each competitor on the same framework. Identify where you lead and where you're vulnerable.

Step 5 — Build the moat investment roadmap

  • Next 90 days: Widen the primary moat
  • Next 6 months: Activate the second moat
  • Next 12 months: Defend against competitive threats
  • What NOT to invest in: Moat types scored 0 for both current and potential
  • Quality check before delivering

    Every moat claim is specific — not "we have better AI"
    Current score is honest — not aspirational
    Competitor moats assessed — not just your own
    Roadmap includes "what not to invest in"
    12-month milestone defined for primary moat
    Suggested next step: Stress-test every investment: "If our top competitor prioritised this for one sprint, could they match it?" If yes, it's not a moat — it's table stakes.