The Decision
Verdict: SERVICES-FIRST, with Australia as the preferred first beachhead. Build this first as a productised commercial-recovery service with software underneath, not as standalone venture SaaS.
Three folded decisions: what kind of company, where it starts, how much to fund.
Services-first, not SaaS. Pure SaaS dies early here, for three reasons:
- The field-capture wedge is already commoditised — the buyer won’t pay serious money for another capture UI.
- The highest-value output is judgement-heavy (entitlement, causation, notice compliance, quantum) and needs commercial/legal review. “An unreviewed AI answer is not a product; it is a liability generator.”
- The moat depends on data permissions buyers won’t grant a cold signup.
First vehicle = productised recovery service: fixed diagnostic fee + account retainer + optional success-linked upside, with software underneath that automates as repeatability is proven. Honest label: “a strong specialist services-first company with a possible SaaS/data moat later… not yet a venture SaaS company.”
Australia-first. The de-biasing scan killed the “empty market” pitch in every geo, but produced a ranking. Criteria weighted 30% willingness-to-pay / 30% recovery regime / 20% competitive density / 20% cold-start ease:
| Geo | Score |
|---|---|
| Australia | 4.35 |
| UK | 4.05 |
| US | 3.55 |
| Canada | 3.35 |
AU wins on its statutory Security of Payment regime → the cleanest “commercial event → payment claim/adjudication → recovered cash” loop of any English-speaking market. Half the capture loop is legally mandated; rapid adjudication turns substantiated events into near-term cash, not multi-year litigation; public state adjudication datasets (Victoria, Queensland, NSW) are ready-made fuel for the cross-firm benchmark others keep private.
- US — biggest prize, highest ACV, worst FIRST market: Magra already there, platforms denser, no dedicated QS buyer.
- UK — natural second, not empty: Gather Insights owns the front of the NEC/JCT motion.
Detail: geo comparison, Australia, US, Canada.
Fund Stage 0 only. Not “raise and build for a year.” Fund exactly one 30-day stage — beachhead validation — and let real evidence release the next tranche. Stage 0 gate: 20+ senior AU buyer interviews, 5+ firms providing redacted project packs, 3+ agreeing paid diagnostic terms in principle, AU counsel confirming the service boundary is legal. Fail (no paid intent / no records access / can only operate as a regulated legal practice) → stop cheaply.
Hypothesis chain
Each hypothesis has a validate-and-advance or fail-and-exit gate.
| Hypothesis | What it claims | Gate |
|---|---|---|
| H1 | Real demand for recovery exists | Validate or exit |
| H2 | Firms hand over the messy data | Validate or exit |
| H3 | Diagnostic produces genuine value (recovered cash, not nice summaries) | Validate or exit |
| H4 | Work compounds into reusable structured outcomes | Validate or exit |
| H5 | Model generalises beyond bespoke archaeology | Validate or exit |
| H6 | Cross-firm data loop becomes a real moat | Validate or exit |
| H7 | Model travels to next jurisdiction without a rebuild | Validate or exit |
Concrete kill criterion: if “fewer than 3 of the first 5 paid pilots allow your pack into a live payment claim, adjudication prep, final account or negotiated recovery,” the thesis is failing — stop or reshape. Full staged plan: entry plan.
Corrected positioning vs the seed call
- Not the voice diary — now the capture front door.
- Not an empty market — recovery is contested, and the strategy says so.
- Not self-serve AI — a service with a human in the loop on every contractual output.
- Not the biggest market — the one where evidence-to-cash is fastest and the moat data is uniquely bootstrappable.
Same sentence the work started circling — capture what happened, prove why it matters, turn it into commercial action, reuse the history to price the next job — now with a country, a vehicle, a buyer, a moat, and the exact list of things that would prove it wrong.