Founding session handoff (9 Oct 2026)

Purpose: let a fresh agent pick up this project with the full founding context. Read this plus README.md and CLAUDE.md before making changes. The raw source is 2026-10-09-francois-call-transcript.md (same folder).

Session narrative

1. Stevan pasted the Granola transcript of his catch-up call with Francois deCourtivron (retained carve-out CFO for Accel-KKR) and asked what it teaches about CT's partnership program and what to do.
2. Claude produced the analysis (distilled below), saved partner intel to global memory (ct-francois-akkr), and logged follow-ups in stevan-open-loops.
3. Stevan then asked for this project: a log of everything needed to reach the north star ("most partnerable business in the payments space"), seeded with the analysis as a ratable plan, on a live pages.dev board. He added two of his own ideas from during the call: a partner capabilities bot on WhatsApp (possibly opening up the Compass Ask Slack bot) and partner account self-serve on WhatsApp (earnings, referral code, no portal login). These became items T-02 and T-03.
4. The board was built, deployed, and QA'd end to end the same day (see "State" below).

The strategic reads (the why behind the items)

These conclusions are the reasoning layer that data/items.json compresses; don't lose them when editing items.

1. Francois is a channel, not a referrer. He serially embeds in freshly carved-out companies at the exact moment treasury and banking are chosen from scratch. AKKR: ~$23B AUM, ~70 portfolio companies, $50-250M revenue, always acquiring. One operator = a pipeline of companies. This is a new partner archetype ("fiduciary operator") the program wasn't built for → items P-02, G-01.
2. The first trade is the product. At Accertify ("Certify" in the transcript), a PE-introduced client's first trade of $20-25M stuck for 3 days on a missing purpose code. Francois's verbatim fix: "white glove service... that first trade, that first, second trade... Did you put a purpose code in? Did you know the bank has to stamp this?" The window where the segment is won or lost is trades 1-2 → items P-01, T-01, C-02, C-04.
3. Speed-to-quote closes. Nico live-quoted a $5M trade on the phone, showed ~$50k saving vs the bank's quote, and it "went straight to the CFO". For US corporates the whole sale can be one phone call, which needs quote-capable coverage at 4pm ET → items P-06, C-03.
4. The ICP is repeatable. US-HQ'd PE portfolio companies with small international subsidiaries: bulge-bracket primary banking ("we always get screwed on exchange rates, always"), FX never top-3 but acknowledged waste, needs episodic and urgent ("windows"). Route in via the operator class, not marketing to CFOs → items G-01, G-02, C-06.
5. Watch-outs. (a) Francois clocked "Corpay" instantly; "whatever Corpay can do we can do, plus rev share" invites "why not go direct?": always frame multi-provider best execution (Corpay AND Currencycloud; CT can offer named accounts via both, see memory ct-named-accounts) plus the service layer → item S-01. (b) The call opened with a partner payment Francois had chased; payout reliability is program hygiene → P-04. (c) His funding question was counterparty diligence; bootstrapped-profitable-since-2018 is a sellable stability point for PE audiences → folded into C-01. (d) He must stay neutral (discloses, offers to waive comp); salesy co-branded collateral is unusable by him: he needs procurement-grade neutral material → C-01, P-02.

Commitments Stevan made on the call

• Add Francois on WhatsApp (Stevan's number was pasted in the Teams chat; Stevan messages first).
• Send collateral as a live link, not stale PDFs, plus a partner portal login.
• Share partner portal walkthrough videos for feedback over time.
• Senior escalation on all Francois introductions (Paul or Stevan, not the default queue).

Live opportunity

Francois's current carve-out ("Staratas", sp?): US→MYR payroll funding for ~70 Malaysia staff, JPM banking, high rates, "not top of the list but it's there", growing; future non-USD receivables (GBP etc.). Play: C-05 (MYR one-pager + offer to benchmark his recent MYR confirmations). Confirm CT's actual MYR capability with Paul before sending anything.

State at save time (9 Oct 2026, end of founding session)

• Board live at https://ct-partnerable-lk0q-om10-j6jo.pages.dev with 22 items: 19 proposed/unrated, 3 in flight (T-04, T-05, G-03). KV ratings store empty; Stevan has not rated yet.
• QA passed in a real browser: live sync, rating round-trip to KV and clear, mobile layout, zero JS errors.
• Repo committed (3 commits), not pushed anywhere; devbox-local.
• Nothing has been sent to Francois yet: no WhatsApp message, no pack, no MYR note.

Where everything lives

• Loop mechanics, infra IDs, item schema: README.md. Session rules: CLAUDE.md.
• The plan itself: data/items.json (+ data/ratings.json once synced).
• Global memory (machine-level, richer partner intel): ct-francois-akkr, ct-partnerable-board, ct-named-accounts, ct-ddre-partnership, stevan-open-loops.
• Raw call source: docs/context/2026-10-09-francois-call-transcript.md.
• Deep CT strategy/wiki: ares (mirror: ssh hermes ~/wiki).

Operational gotchas learned this session

• Cloudflare Pages first-deploy: canonical URL can throw 522/523 for the first minute or two; retry before concluding failure. Share the canonical *.pages.dev URL, not the per-deploy hash URL.
• gstack's browse toolkit is NOT installed on devbox (skill stubs only). For board QA use playwright-core from an existing project, e.g. require('/home/stevan/dev/ela-wt-family-mint/node_modules/playwright-core') with its cached Chromium.
• Google Fonts woff2 URLs come as /l/font?kit=... (no .woff2 suffix); extract the URL from the /* latin */ block of the css2 response with a browser User-Agent.

Suggested skills for follow-on sessions

• frontend-design + design-taste-frontend: required by Stevan's global rules before touching the board's HTML/CSS/JS (and dataviz if adding charts/tiles).
• browse (gstack): QA after redeploys, if/once installed on this box; otherwise the playwright fallback above.
• copywriting / direct-response-copy / humanizer: drafting C-01 (vendor pack), C-05 (MYR one-pager), S-01 (positioning paragraph).
• ct-sql / ct-pd: partner, revenue, or corridor data pulls to size items (e.g. first-trade stuck rate baseline for P-01).
• otto-plan / otto-orchestrate: when T-02/T-03 (WhatsApp bots) graduate to builds; the Compass Ask code lives around ~/dev/compass-ask.
• jira-style: if any item becomes a CT Jira ticket (e.g. T-01 corridor validation for the dev team).

How to resume

1. node scripts/sync-ratings.mjs to pull Stevan's ratings; if rated, regroup the plan around his buckets (his rating beats suggested).
2. Check stevan-open-loops memory for which personal actions (WhatsApp, pack, MYR note) are still outstanding.
3. New partnership insights from any future conversation get added as items and deployed, per CLAUDE.md.