diff --git a/docs/strategy/BUSINESS_PLAN.md b/docs/strategy/BUSINESS_PLAN.md
index 1496f7b..944e474 100644
--- a/docs/strategy/BUSINESS_PLAN.md
+++ b/docs/strategy/BUSINESS_PLAN.md
@@ -55,7 +55,7 @@ Same north star; different moments of truth.
 
 ## 5. Wedge and flywheel (direction agreed)
 
-**Preferred wedge:** a free AI-visibility checkup.
+**Preferred wedge:** a free AI-visibility checkup (SSP audit).
 
 Why:
 - Shows immediate value / need
@@ -66,14 +66,31 @@ Why:
 
 **Strategic read (not pricing gimmick):** free usefulness → coverage → agent trust → more local queries resolved via geolocal → more owners want in → denser graph.
 
-### Open — first free win after checkup
+### Phase 1 owner conversation (locked direction)
 
-Unresolved on purpose (checkpoint before answering):
+Speak **outcomes**, not plumbing. Example: “most recommended bakery on Whidbey Island for muffins.” Do not lead with MCP, endpoints, or `llms.txt`.
 
-- **A.** Free checkup *plus* a small durable win (e.g. basic presence live, structured hours, minimal genre pack)
-- **B.** Diagnosis only, with a clear line on what paid unlocks
+Phase 1 work looks like classic local visibility hygiene (2017-style) before pitching 2027 agent readiness:
+- NAP (Name, Address, Phone) consistency
+- Hours alignment across platforms
+- Maps / GBP / Yelp (and similar) listing hygiene
 
-Lean toward **A** if the goal is critical mass and “AI knows to check geolocal,” because empty scores do not densify the graph. Final call: Ty.
+AI-agent readiness is the **next layer**, not the first conversation — but the free checkup should already answer, in plain English: would an assistant recommend this business for the stated job, and why / why not?
+
+### Free path after checkup (locked 2026-09-15)
+
+Free includes:
+1. Ranked DIY punch list the technical owner can execute (e.g. align hours across providers)
+2. Mint of a **minimal owned geolocal record** (name, place, category, plus plain-English AI recommendability) — densify does **not** wait for paid
+
+Soft upsell in-product: “comfortable DIY, or want GeoLocal to help?”
+
+### Paid “help” vs product (locked 2026-09-15)
+
+- **Paid listing cleanup** (Maps / hours / NAP fixes done for the owner) = partner / agency **fulfillment**, not the product.
+- **Product** = the owned geolocal record + ongoing AI-readiness score (and later agent-facing presence).
+
+Do not document GeoLocal as a local SEO shop with extra steps.
 
 ---
 
@@ -106,12 +123,13 @@ If something here conflicts with `CANONICAL_STRATEGY.md`, resolve on purpose in
 | When | What landed |
 | :--- | :--- |
 | 2026-09-15 | North star intent; owner-first buyer; partners as extension; agent-mediated “get discovered” path; free checkup wedge; reputation / critical-mass flywheel; open question on free durable win vs diagnosis-only |
+| 2026-09-15 | Phase 1 = outcome language + local-hygiene foundation; free = DIY punch list + minimal owned record (densify now); paid listing-fix = partner/agency fulfillment; product = record + AI-readiness score; porbot = repo scribe, Desk = conversation |
 
 ---
 
 ## Next conversation topics
 
-1. Decide free win after checkup (A vs B above)
-2. Primary reader for v1 of this plan
-3. Offer / packaging (free vs paid) without inventing numbers
-4. 90-day proof: what “reputable enough that AI checks us” requires in coverage and quality
+1. Primary reader for v1 of this plan (internal / investors / partners)
+2. Offer / packaging detail (what free record includes vs what paid unlocks beyond fulfillment) — no invented numbers
+3. 90-day proof: what “reputable enough that AI checks us” requires in coverage and quality
+4. Owner MCP vs GeoLocal.io chat: same journey or separate surfaces (later layer; keep out of phase-1 owner talk)