Articulate · Internal Decision Document

The build-partner opportunity

Fintan is assembling a claims-generation venture — AI engagement engines deployed against partner databases, vertical by vertical — and recruiting Anthony as the build partner. This page lays out the roles, phases, costs, time, and the criteria that make it a yes or a no.

10 June 2026 · sources: three Fintan voice notes (16:14 / 16:56 / 17:28) · market dossier · competitor database

£15

per clean claim to Anthony (Dynamo, agreed split per note #3)

20,000

warm energy claimants to re-sign — pilot-shaped, available now

verticals on the table today (PCP · business energy); more behind

~Nov 26

Tribunal gate on the PCP scheme — the clock the whole play runs on

01What the venture actually is

Database owners (insurers, introducers) hold millions of consented-ish customer records they can't monetise — calling them is too expensive. Fintan's network brings those databases plus a regulated claims chain (LEAD SOCIAL LTD, FCA FRN 951156 → Blue Lion Law / SRA firms → pcpclaims.info). The missing layer is an AI engagement engine — the Emily/Oliver pattern — that converses with the database at near-zero marginal cost, surfaces eligible claims, and routes them into the regulated chain. Revenue is per signed claim, shared three ways: database owner ~50%, Fintan + Anthony the rest.

02Roles

Seat	Owns	Held by
Deal-maker	Introducer relationships, database deals, commercial terms, the regulated chain	Fintan
Build partner	The engine: conversational AI (WhatsApp/web/voice), vertical knowledge brains (RAG), e-sign LOA flow, funnel integration, compliance-by-design scripts, conversation audit logs, reporting & ops automation	Anthony — the open seat
Database owner	Data, customer consent, brand cover ("Dynamo Dave"), takes ~50% of claim fees	Steve/Dynamo · energy introducer · future partners
Regulated processor	FCA CMC permission + SRA litigation capability; processes and monetises the claims	LEAD SOCIAL (FRN 951156) / Blue Lion Law — verify, step 1

What a 10/10 build partner looks like

Ships in weeks, not quarters — engine assembled from the existing stack (LLM + RAG + WhatsApp Business API + e-sign), not built from scratch.
Compliance is in the code — free-route disclosure, no false urgency, opt-out handling and full conversation logs baked into every script. In a niche under active FCA/SRA investigation, the compliant engine is the durable one. 50 of 70 competitor sites fail this today (measured).
Everything measured — contact → response → qualified → LOA signed → clean case, per vertical, per day. The £15/claim only exists if clean-case rate is provable.
Portable across verticals — same engine, swap the knowledge brain: PCP → energy re-sign → next database. The moat is the speed of porting.
Doesn't become a job — ops automated (escalation queues, anomaly alerts), not a call-centre replacement that consumes Anthony's week.

Gap analysis: items 1, 3, 4, 5 are Anthony's existing stack and instincts. The genuine gaps: a solicitor review loop for scripts (buy: £2–5k), a signed introducer agreement defining the £15 (Fintan's job to produce), and capacity (the real constraint — GIG ≤2 days/week, Roche-Debbie HOT, Articulate Phase 1).

03Phases & timescale

Phase 0 — Diligence · now → ~2 weeks · ~2 days of Anthony-time
Verify FRN 951156 + Blue Lion Law SRA record. See the introducer agreement and fee flow on paper (who pays the £60/£45, when). Energy: see the 20k list's consent basis + the original LOA pack. Kill criteria below — any failure ends it cleanly.
Phase 1 — Energy re-sign pilot · weeks 3–8 · ~12–15 days of Anthony-time
The best-shaped first project of everything pitched: existing clients, warm list, B2B (lighter marketing rules), no Tribunal dependency. Build the engine on the energy knowledge brain, re-sign the 20k via conversational outreach + e-sign LOA. Measured outcome inside 60 days.
Phase 2 — Dynamo PCP · build Sept–Oct, launch on the Tribunal decision (~Nov 2026)
Port the engine: PCP knowledge brain, "Dynamo Dave" persona, reg-plate-first intake (the competitor analysis shows nobody runs conversational intake — open slot). Launch the week the gate opens; the queue position is the prize. Window: ~12–18 months before lender proactive contact erodes it.
Phase 3 — Vertical factory · 2027
Each new database = knowledge brain + persona + commercial agreement on the same engine. Fintan says the introducer has ~5M records behind the 20k. Only enter on Phase 1–2 evidence.

04Costs & time

Build & run costs (Phase 1 pilot)

Engine build: £0 cash — existing stack (Claude/GPT API, RAG, WhatsApp Business API, e-sign). The cost is Anthony-time: ~12–15 days across 6 weeks.
Run cost, 20k conversations: LLM ~£0.05–0.15 + WhatsApp ~£0.04–0.08 per conversation + e-sign ~£0.50 per signed LOA → ≈ £3–5k total at realistic volumes.
Legal/compliance review: £2–5k one-off (script sign-off via the partner firm).
Comparison: human re-sign of 20k at ~£25/contact ≈ £500k. The engine does it for under 1% of that — this is the entire commercial argument, and it's defensible.

Revenue (corrected, not Fintan's headline)

Dynamo PCP: £15/clean claim to Anthony. Corrected envelope 3k–18k claims (dossier §5) → £45k–£270k, one-off harvest, paid as cases sign (CPA) — late 2026 into 2027. Fintan's £375k assumes his 25k; treat as ceiling.
Energy re-sign: fee per re-signed LOA unstated — ask. Even at £10–20/re-sign on 4–6k conversions, £40k–£120k; the downstream £3k-commission claims are litigation-timeline money (years, not months).
Cross-sell: asserted, unmodelled. Ignore for the decision.
Shape: this is a revenue-share venture, not a fee engagement — no Marketing Engine Pilot pricing applies (consistent with the locked partnership frame).

05Success criteria

Gate	Pass	Fail / kill
Phase 0 diligence	FRN + SRA verified · introducer agreement on paper with the £15 flow · energy list has lawful re-contact basis · original LOA pack seen	Any entity unverifiable · fee flow stays verbal · consent basis can't be evidenced → stop, no build
Phase 1 pilot (60 days)	Re-sign conversion ≥20% · cost/re-sign ≤£1.50 · zero ICO/complaint flags · Anthony-time ≤15 days actual	Conversion <10% · compliance incident · ops eating >2 days/week → stop or renegotiate
Phase 2 Dynamo (90 days post-launch)	≥5% claim conversion on the contactable slice · £15/clean claim actually paid under signed agreement · clean-case rate ≥60%	Tribunal guts the scheme · Dynamo DPO can't confirm marketable slice · payment friction → engine survives, redeploy to next vertical

06The honest risk register

Capacity collision (highest likelihood): Roche-Debbie is HOT, GIG is capped, Articulate Phase 1 is the year-1 revenue plan. The pilot's 12–15 days must come from somewhere. Mitigation: Phase 0 is 2 days and decides everything; commit nothing past it.
Counterparty verbosity: three voice notes, zero documents. Every number is verbal. Mitigation: Phase 0 converts talk to paper or kills it.
Regulatory heat: FCA/SRA actively investigating claims-marketing in this exact niche. Mitigation: compliance-by-design is the product, not overhead — and it's also the pitch differentiator.
Tribunal risk (PCP only): scheme amended or delayed ~Nov 2026. Mitigation: energy pilot is Tribunal-independent; the engine is portable by design.
Fintan pattern risk: second venture pitch in three weeks (Dubai.co.uk still un-discovered). Enthusiasm outruns paperwork. Mitigation: same discipline both times — structured gate before build.

Recommendation: say yes to the seat, gated. Reply committing to Phase 0 only: the three documents (introducer agreement, FRN/SRA verification, energy-list consent basis) and the energy re-sign pilot as the first build — not Dynamo, not five verticals. 2 days of diligence buys the decision with evidence instead of voice notes. If Phase 0 passes, the pilot is the cheapest £40k+ test Anthony will run this year and the engine it produces is the productised asset Articulate keeps regardless of how the Fintan relationship evolves.

Articulate internal · not legal or financial advice · companion: market dossier · competitor database · vault: Clients/Fintan/ (voice notes, this page source)