Baseline · July 18, 2026

Where EdgePointe stands vs. the AI frontier

Baseline · State vs the Frontier
Seven cards, ~30 seconds each. You get the gist; the full report carries the detail.
EDGEPOINTE · VANTAGE
Where we stand

Leading where it's hardest to copy — one gap holds us back Real lead

The honest headline: we lead on the two hardest things to copy, and have a gap on the one discipline that would make that lead compound.

  • Leading: our agentic build pipeline and design-as-a-system — most agencies and MSPs have nothing like it.
  • The gap: formal evals — measuring quality so it's enforced, not hoped for.
  • Close that one gap and the lead becomes defensible.
Bottom line: we're ahead where it counts — the work is protecting and compounding it.
The scorecard

Seven domains, four honest grades One gap

One rating per domain — Leading, Solid, Developing, or Gap.

Agentic pipeline
Leading — a sophisticated build harness
Design as a system
Solid → Leading — taste already codified
Evals & measurement
Gap — highest-leverage place to invest
Platform (Cloudflare)
Solid — production-grade multi-tenant
AI product (Vantage)
Leading vs. SMB peers
Business model & moat
Developing — biggest strategic upside
Read it this way: five strong domains, one gap — and the gap is the lever.
The top move

Stand up an evals discipline — do this first Register #3

We have lints, mirror-score, and a visual-QA ritual, but no formal evaluation asset.

  • Build a versioned golden set of 'elite vs. not' pages, seeded from our own past builds and their bugs.
  • Score it with a calibrated critic that gates promotion — quality enforced in CI, not hoped for.
  • Every failure we've already caught is a golden-set case waiting to be filed.

This turns Paul's taste into a compounding, teachable, defensible asset.

Highest return on the board: the move that converts taste into a company asset.
The growth move

Point the same engine at Managed IT Register #8

Our competency isn't cloning venue sites — it's AI-native operations: agentic pipelines, measurable taste, edge deployment.

  • Aim it at Managed IT: automated ticket triage, security monitoring, QBR reports, billing reconciliation.
  • Automation-first MSPs hit 18–22% EBITDA vs. 11–14% for traditional shops.
  • Only 13% of MSPs monetize AI as a revenue line — while 48% say clients will demand it.

Same competency, new revenue lines — growth that's exponential, not linear.

The anti one-trick-pony move: lift margin on the book we already have.
The thesis

The moat isn't the model — it's the flywheel The bet

In 2026, model access is not a moat. The durable moats are workflow, proprietary data, distribution, and brand — and we hold four of five.

  • We're not betting Claude stays ahead of GPT or Gemini. We're betting on whoever encodes their taste, workflow, and client data into an AI-native system first.
  • A cheaper model each quarter is a tailwind — our cost to produce drops while the moat holds.
  • Next step: make the data flywheel explicit — every build, edit, and ticket feeds back in.
Say it plainly to Bob and Niven: the model is a commodity input; the flywheel is the moat.
Bottom line · next steps

Five strong domains, one gap, two first turns Starting line

This baseline is the starting line; future updates are deltas against it.

  • Do first: stand up the evals golden set (Register #3).
  • Then: point the engine at Managed IT delivery (Register #8) — the flywheel's first two turns.
  • Track: everything routes through one ranked backlog — the Opportunity Register.

Cadence: daily watch, weekly digest, monthly capability review, quarterly moat review with leadership.

Kept current, one file is the growth engine: the Opportunity Register.
Card 1 / 7 · ~20s
⬇ Full report (PDF)