lighthouse

name

lighthouse

description

Lighthouse-buyer workflow — find and prepare the first paying government buyer for the SE evidence + tender-pack product. Mines se_buyer_prospects (buyers who already contract with registry SEs), shortlists VIC/SA buyers with social-procurement obligations, and assembles an evidence-led outreach pack. Use when the user says "/lighthouse", "find buyers", "prospect buyers", or wants outreach material for a procurement team.

Lighthouse Buyer Workflow

Goal (buyer-wedge move 3): one Vic or SA government buyer using a tender-pack in a real procurement. The pitch is their own data: "your agency already bought $X from N social/Indigenous enterprises — here is the evidence, and here is the tool that makes your next tender's social-procurement weighting defensible."

Stage 1 — Refresh prospects

node --env-file=.env scripts/scout-se-buyers.mjs            # dry run, top 20
node --env-file=.env scripts/scout-se-buyers.mjs --apply    # rebuild se_buyer_prospects

Table: se_buyer_prospects (buyer_name, se_supplier_count, contract_count, total_value, last_contract_end, certified_supplier_count, example_suppliers, states).

Stage 2 — Shortlist

Data reality (verified 2026-06-08): the prospect pool is FEDERAL — AusTender is commonwealth procurement; VIC/SA SPF/SAIPP obligation-holders tender through state systems we don't ingest yet. Top of pool: Defence (255 SE suppliers, $2.3B), Services Australia (174), PM&C (145), DSS (137), NIAA (132).

Prioritise in this order:

1. Federal buyers via the IPP angle — the Indigenous Procurement Policy sets MANDATORY targets for commonwealth buyers, and ~80% of the registry is Supply Nation/ORIC. The pitch: "here is your agency's IPP delivery story, per-supplier, evidence-linked" — no one else can produce it. Strongest combination of obligation + our evidence.
2. High se_supplier_count + recent last_contract_end (active buyers, not historical).
3. certified_supplier_count > 0 (their suppliers carry marks a probity advisor recognises).
4. VIC/SA buyers (SPF/SAIPP obligation) — currently NO contract evidence in our DB. Unblocking them needs a Buying-for-Victoria / SA Tenders ingest — raise with the user as a /wedge question before building (it's evidence-depth work, not widening, but it's still a build).

Cross-reference: which shortlisted buyers also appear as policy targets in apps/web/src/lib/social-procurement.ts (state policy inserts).

Stage 3 — Outreach pack (per prospect)

Assemble, do not send (sending is Tier 3 — human does it):

1. Evidence one-pager — their SE spend story from our data: total value, supplier count, named examples (from example_suppliers), with per-claim provenance. Every dollar figure needs a source row.
2. Live tender-pack for a plausible upcoming category for that buyer (/procurement/tender-pack), state policy citations included.
3. Draft email — short, evidence-first, no marketing voice. Lead with their number ("Your agency contracted $X with N social enterprises since YYYY"). One ask: 30 minutes to show the tender-pack against a live procurement.
4. Save pack to thoughts/shared/prospects/<buyer-slug>/ (one-pager.md, email-draft.md, notes.md).

Stage 4 — Track

Keep a simple ledger at thoughts/shared/prospects/PIPELINE.md: buyer, stage (identified → pack built → contacted → meeting → live tender → paying), date, next action. Update it every time this skill runs.

Constraints

• NEVER send outreach from a session (Tier 3 — drafts only, human sends).
• Every claim in an outreach pack must trace to a queryable row — provenance discipline applies doubly to material leaving the building.
• Don't spray: depth on 3-5 prospects beats packs for 20. The goal is ONE live procurement.