Construction-tech strategy · internal
One bet, stated as a hypothesis
Services-first · Australia-first · fund Stage 0 only
Hypothesis: the defensible opening in construction AI is a commercial-recovery service — turn site evidence into recovered money, sell it as a service before SaaS, start where the evidence-to-cash loop is fastest. This is the evidence trail behind that call: the goal, the pivot, the thesis, ~60 competitors scored, and the gates that would prove it wrong.
Executive summary
From a list of pain points to one falsifiable bet
A contractor call surfaced seven near-term AI use cases. Instead of building the seed's own favourite — a voice site diary — we mapped the whole field, reframed off the crowded capture layer onto the commercial event, scored ~60 competitors on one rig, and converged across two independent reviews on a services-first, Australia-first commercial-recovery play, funded one 30-day stage at a time.
1 The seed
A call with a contractor surfaced seven candidate AI use cases — site diaries, tender extraction, cost planning, programme dashboards, O&M manuals, drawing takeoff, materials tracking. Its own recommendation: start with the voice diary.
All seven use cases →2 The reframe
"Voice note → pretty PDF" is a commoditised knife-fight — Raken, Fieldwire, PlanRadar and Procore's own AI agents already ship it. So the diary became the capture mechanism, not the product. The product is the layer above: capture → evidence → commercial action → reusable cost memory. Sell recovery, not admin.
Why we pivoted →3 So we researched the field first
The market-research brief, in order:
- Find the players in the space.
- What they do well — the problem each solves, the pitch vs. what they actually deliver.
- Both sides of the UX — the field-entry experience and the management dashboard.
- How much AI can build on top — API quality and openness.
- Team size and funding — how much firepower sits behind each.
- Mine the reviews — Capterra and real user data: what users actually like and hate.
- Where the gap is for us to play.
Every company was then scored on the same rig: a 21-area coverage grid (where a tool plays) plus a 4-axis opportunity lens (whether and how we can attack), reviews bias-tagged so vendor-solicited 5-stars were down-weighted. ~60 competitors, identical scales.
4 What we ran — the index
5 The play, and why
| Services-first | Pure SaaS dies early — capture is commoditised, the high-value output needs legal review, and the moat needs data a cold signup won't grant. Ship a productised recovery service; automate underneath as it repeats. |
|---|---|
| Australia-first | Statutory Security of Payment gives the fastest event → claim → cash loop of any English-speaking market, and state adjudication data is ready-made moat fuel. Ranking: AU 4.35 · UK 4.05 · US 3.55 · CA 3.35. |
| Fund Stage 0 | One 30-day validation gate — 20+ buyer interviews, 5+ data packs, 3+ paid-in-principle, legal boundary confirmed. Evidence releases the next tranche, not a 12-month build. |
The empty space is precise: no scaled, trusted, jurisdiction-specific recovery platform with cross-firm outcome data. Recovery itself is contested — Magra and a 2024–25 micro-cohort exist — but cross-firm cost is empty by design, and that is the moat.
The argument, in eight steps
- 01 The Goal
- 02 The Starting Menu
- 03 The Pivot
- 04 Why Not Everyone
- 05 The Thesis
- 06 The Research
- 07 The Landscape
- 08 The Decision
Short on time? Jump straight to the decision →