2026-07-30 — Knowledge base seeded from the GEO Agency Playbook
`JSGlobtech_GEO_Agency_Playbook.pdf` — the confidential business blueprint authored 2026-07-27 — was copied into `knowledge/source/` and extracted into the knowledge base. The play...
2026-07-30 — Knowledge base seeded from the GEO Agency Playbook
Hand-written entry, not produced by
scripts/kb-sync.sh. It records a documentation event rather than a sync window, so it is named by topic instead of by SHA and does not touchknowledge/.kb-state. The sync-generated draft for the same date is2026-07-30-ead6c62.mdand covers the code history.
What happened
JSGlobtech_GEO_Agency_Playbook.pdf — the confidential business blueprint authored 2026-07-27 — was copied into knowledge/source/ and extracted into the knowledge base. The playbook is roughly 85% go-to-market and 15% technical, so two new sections were added alongside the existing engineering docs:
strategy/— positioning (SEO vs GEO, the Agentic Action Layer), the four target UAE verticals, and reporting KPIs.sales/— pricing tiers, the non-technical pitch script, both cold outreach templates, and the two-step live demo framework.
The existing engineering docs were enriched rather than replaced: architecture/geo-stack.md, integrations/mcp-server.md, and integrations/lead-capture.md now document the llms.txt layer, the MCP endpoint, and lead delivery in detail, each verified against the code. operations/client-engagement-sop.md captures the outreach → demo → close loop and the recurring obligations each sale creates.
Seven decision records (ADR-0001 to ADR-0007) were written for choices the playbook and the codebase imply. Six are Accepted because the repository already implements them; ADR-0005 is Proposed.
Why it was worth doing
The playbook existed only as a PDF in a Downloads folder. Everything commercially load-bearing — who to sell to, what to charge, what to say, what the demo proves — lived outside the repo, while the code implementing it lived inside with no record of the reasoning.
The more valuable outcome was the cross-check. Reading the playbook against the actual tree surfaced several places where the sales narrative promises more than the code delivers, and those gaps are now written down next to the claims instead of being discovered in front of a client.
Divergences found between the playbook and the repo
| Playbook claim | Reality in this repo |
|---|---|
| "book consultations", "process reservations" inside chat (§1) | Lead capture only; no scheduling, booking, or payment |
| Tier 3 "Multi-endpoint MCP tools" | Exactly one tool (request_dev_consultation); OpenAPI exposes one path |
| Tier 3 "Live Inventory/Property sync" | No database, ORM, or external data integration |
| Tier 2 "24/7 Hosting & Monitoring" | No monitoring, alerting, uptime checks, or CI |
| "CRM/WhatsApp Lead Integration" | One generic outbound webhook to Apps Script → Sheets; WhatsApp is not a wired channel |
| Monthly citation reports (all tiers) | No tooling and no defined methodology |
| Demo shows a "Confirm / Submit button" (§6) | Host-client UI, not JSGlobtech code; the tool returns JSON |
| "Over 40%" of clients skip Google (§4, §5) | Unsourced; flagged for verification in strategy/kpis.md |
Also corrected: architecture/README.md described lead_id as a bare UUID; it is actually jsg-<uuid>.
⚠ Unresolved — needs owner decision
The playbook's pricing tiers ($3,000 / $6,000 / $15,000+ setup with $1,000 / $2,000 / $4,000+ retainers) contradict the published public/.well-known/llms.txt and public/llms.txt, which describe GEO/MCP as "Scoped per engagement (often bundled with app/web builds)" and lead with Flutter from $10,000 and web from $5,000.
The contradiction is public and machine-readable — a prospect or ChatGPT can read the site's version. Tracked in decisions/0005-productized-tiers-vs-custom-scoping.md (Proposed) and flagged at the top of sales/pricing-tiers.md. No quote should go out using the tier numbers until this is settled.
Follow-ups
- Resolve the pricing contradiction (ADR-0005) and make
lib/content.ts, bothllms.txtfiles, andGeoSchema.tsxagree - Find a citation for the "over 40%" statistic, or soften the wording in the pitch script and email template 1
- Fill the
[Your Phone Number]placeholder before any outreach is sent - Write a citation-report methodology — it is the recurring deliverable on every tier and has no defined process
- Add monitoring before selling a Tier 2 "24/7" commitment
- Add a check that
public/llms.txtandpublic/.well-known/llms.txtstay identical - Write outreach templates for the real estate and dental verticals