Vocap
← Back to Insights
Emery Waddell

The Four-Part Moat: Revisited, Part 1

2026-11-10

Part 1 of 3: Vertical focus is the wedge, not the moat

The short version

  • In March we laid out a four-part moat for vertical AI: vertical focus, embedded best practices, cornered data and system of action.
  • Eight months later we would draw the lines differently. Vertical focus is the wedge. It gets a company in the door ahead of horizontal vendors. It is not what keeps competitors out.
  • Industry understanding, which we called the heart of the moat in March, no longer protects anyone on its own. Foundation models absorb more of it with every release.
  • Deep integration and compliance know-how are the path from wedge to moat, because they are what earns a product the right to act.
  • Cornered data and system of action are the moat. We sharpen our definition of each in the next two posts.

What we said in March

In Part I we listed brand trust, domain knowledge, channel partnerships, legacy integrations and last-mile product depth as the sources of durable protection. In Part II we went further and called "semantic understanding" the heart of the moat: knowing how an industry's entities relate, when a recommendation is confident enough to act on, and which mistakes are catastrophic versus merely annoying. We said that took years of iteration with real users and could not be shortcut.

What has changed

Industry knowledge is still valuable. It is no longer exclusive.

Founders who come from an industry bring that understanding with them, and it still shapes a better product. What has changed is that it is no longer scarce. Every model release absorbs more of what has been written about a field, including highly technical ones. Take commercial real estate underwriting. Hand a frontier model a rent roll, a trailing twelve month operating statement and an offering memo, and with little or no tuning it will map the line items into a standard model, flag the big lease expiring in year two, and understand that overstating net operating income is far more dangerous than a rounding error in repairs. That used to take a purpose-built product. Now a workflow vendor can build on the same model and show up in your vertical with something that looks comparable on the surface.

Vertical focus is the wedge

Industry understanding is becoming table stakes, but vertical focus still gets a company in the door ahead of any horizontal vendor.

  • Go-to-market. Industries are tight communities. References, word of mouth and a few key conferences do most of the selling, and a focused company knows exactly whom it is selling to.
  • Last-mile experience. A product built around how people in the industry actually work, in their language, beats a general tool where the work gets done.

Those advantages are real, and they are how a vertical company lands its first customers. They do not stop a well-funded competitor who does the same work two years later. That is the definition of a wedge, and it is why we now ask about the path as closely as the current position.

Integration and compliance: the path from wedge to moat

Two of the advantages we listed in March are more than wedge. They are the preconditions for the right to act.

  • Integration depth. Vertical companies do the tedious work of connecting to industry systems of record like Yardi, Procore, Epic or a TMS. Read and write access to those systems, many without modern APIs, is what makes everything else possible.
  • Regulatory know-how. Certifications, audit trails and liability rules differ by industry. They are a barrier to entry and part of how a product earns the right to act.

A product cannot become a system of action without write access to the system of record, and in a regulated industry it cannot be trusted with consequential decisions without the audit trail and the certifications. This is the work that turns a wedge into a path.

The question we now ask founders

If a frontier model could read everything ever written about your industry, what would you still have that it doesn't?

The best answers come down to two things: knowledge the model can never read, and permission to do the work rather than describe it. Those are cornered data and system of action, the two parts of the original four we would still call a moat. The next post sharpens our definition of each.

Further reading

From Vocap: Forget the SaaSpocalypse. Here's How Vertical AI Actually Wins (the threats behind the selloff, and why data plumbing and pricing come first); Part II: Building the Moat (the original four-part moat).

Elsewhere: Bessemer Venture Partners, Building Vertical AI: An early stage playbook for founders.