Who this is for. The AI CMO reads this map to execute the substitutions when Lisa's seven answers land. Corrina reads it to know which review packs will land in the WhatsApp group next. Laurence, Lisa, and Carla read section 2 (the master matrix) and section 4 (the compliance guardrails).
Companions. /lisa-voice-brief holds the seven questions. This page holds where the seven answers go. /edit-sheet is the Sprint 55.1 side-by-side page where Lisa reads AI-drafted answers on the left and corrects them on the right, one row per question — the read-and-correct alternative to answering cold.
Parent. Sprint 54's /marketing-escalation-map for the domain-line assignment and quorum rules; /marketing-operators-manual for the shared-gate procedure; /ai-cmo-charter for the AI CMO's operating role.
Compliance envelope. Every substitution preserves: ABN, ACN, ACL 387398, mediation disclosure sentence, complaints and privacy links, Y1 audited numbers verbatim, outcome caveat, dignity floor. No beneficiaries are named or identified in any answer or any resulting donor paragraph.
Contents
- 1. The principle — surgical substitution, not rewrite
- 2. Master matrix — the 7 answers x 7 pages grid
- 3. Per-page substitution plans
- 4. Compliance guardrails on every substitution
- 5. Quorum forecast — how many review packs Sprint 55 produces
- 6. Unwind — how to revert a substitution if a human vetoes
- 7. Change control
1. The principle — surgical substitution, not rewrite
Room D already contains ~120 kB of donor copy across seven artefacts. That copy is structurally sound: the compliance markers are in place, the Y1 numbers are correct, the dignity floor is honoured, the frameworks are named. What it is missing is Lisa's voice — specifically, in the paragraphs that make the reader decide to trust LWD versus another charity.
Sprint 55 does not rewrite the seven pages. It swaps about thirty specific paragraphs — the ones where charity-generic voice is currently doing the work Lisa's voice should be doing. Every other paragraph stays as-is. This is why the sprint is small, fast, and safe: the compliance envelope is untouched, the structure is untouched, only the voice-carrying paragraphs move.
What "surgical" means precisely. For each of the ~30 target paragraphs, the AI CMO produces a review pack containing: the exact current paragraph (before), the proposed paragraph assembled from Lisa's answer (after), the source Q number, the domain lines triggered, the quorum level, and the reversibility note. Every substitution is an atomic proposal. Any human can veto any single substitution without affecting the others. Sprint 55 is thirty small decisions, not one big one.
2. Master matrix — the 7 answers × 7 pages grid
Read down the columns: for each page, which of Lisa's seven answers feeds substitutions there. Read across the rows: for each of Lisa's answers, which pages it lands in. A cell with a dot means "yes, this answer feeds a substitution on this page"; blank means "not this page."
| Lisa's answer ↓ · Page → | /donor-one-pager | /donor-stewardship-plan | /foundations-grant-briefing | /foundations-outreach-pack | /founding-donor-deck | /give | /give-sprint48-brief |
|---|---|---|---|---|---|---|---|
| Q1 Why LWD exists (founder story) | ● | ● | ● | ● | |||
| Q2 The moment a trustee should picture | ● | ● | ● | ||||
| Q3 What mediation actually does | ● | ● | ● | ● | ● | ● | |
| Q4 Why 31 households, not more | ● | ● | ● | ||||
| Q5 The wrong reason to give | ● | ● | ● | ||||
| Q6 What year two looks like if it works | ● | ● | ● | ● | |||
| Q7 Why trust you with this | ● | ● | ● | ● |
Read: Q3 (what mediation does) is the most-reused answer — it feeds substitutions on all six externally-facing pages. Q5 (the wrong reason to give) is the most-narrow — it lands only where LWD directly names the donor register it isn't. Q7 (why trust you) is the anchor answer for the four highest-stakes pages.
3. Per-page substitution plans
For each of the seven donor pages, the plan below names the paragraphs that receive substitutions, which Lisa answer feeds each, and the domain lines that will trigger.
/donor-one-pager
| Paragraph target | Source Q | Domain lines triggered | Quorum |
|---|---|---|---|
| Opener paragraph ("Why LWD exists") | Q1 | Philanthropic Marketing-Craft | Domain-touching |
| Paragraph 2 ("Our Y1 approach") | Q4 | Philanthropic Financial & Regulatory Marketing-Craft | Multi-domain |
| "What mediation does" sidebar | Q3 | Philanthropic Financial & Regulatory Marketing-Craft | Multi-domain |
| Footer paragraph ("Why trust") | Q7 | Evidence & Brand Philanthropic Marketing-Craft | Multi-domain |
/donor-stewardship-plan
| Paragraph target | Source Q | Domain lines triggered | Quorum |
|---|---|---|---|
| "Who this is for" opening ("Who our donor is and is not") | Q5 | Philanthropic Marketing-Craft | Domain-touching |
| Forward-look paragraph ("Where stewardship goes in Y2") | Q6 | Philanthropic Financial & Regulatory Marketing-Craft | Multi-domain |
/foundations-grant-briefing
| Paragraph target | Source Q | Domain lines triggered | Quorum |
|---|---|---|---|
| Anchor paragraph ("The moment we're describing") | Q2 | Philanthropic Evidence & Brand Marketing-Craft | Multi-domain |
| "What LWD does" paragraph (next to the ACL 387398 disclosure) | Q3 | Philanthropic Financial & Regulatory Marketing-Craft | Multi-domain |
| "Our Y1 approach" section (why 31, not more) | Q4 | Philanthropic Financial & Regulatory Marketing-Craft | Multi-domain |
| "What year two looks like" forward-look | Q6 | Philanthropic Financial & Regulatory Marketing-Craft | Multi-domain |
| "Why us" closing section | Q7 | Evidence & Brand Philanthropic Marketing-Craft | Multi-domain |
/foundations-outreach-pack
| Paragraph target | Source Q | Domain lines triggered | Quorum |
|---|---|---|---|
| Section 1 opener ("Who we are") | Q1 | Philanthropic Marketing-Craft | Domain-touching |
| "Why this" section (anchor image) | Q2 | Philanthropic Evidence & Brand Marketing-Craft | Multi-domain |
| "What we do" paragraph | Q3 | Philanthropic Financial & Regulatory Marketing-Craft | Multi-domain |
| Section 5 ("Why us") | Q7 | Evidence & Brand Philanthropic Marketing-Craft | Multi-domain |
/founding-donor-deck
| Slide target | Source Q | Domain lines triggered | Quorum |
|---|---|---|---|
| Slide 2 ("Why LWD exists") | Q1 | Philanthropic Marketing-Craft | Domain-touching |
| Slide 3 (anchor image) | Q2 | Philanthropic Evidence & Brand Marketing-Craft | Multi-domain |
| Slide 3b (mediation explainer sidebar) | Q3 | Philanthropic Financial & Regulatory Marketing-Craft | Multi-domain |
| Slide 4 ("Our Y1 pilot") | Q4 | Philanthropic Financial & Regulatory Marketing-Craft | Multi-domain |
| Slide 5 ("Who our donor is") | Q5 | Philanthropic Marketing-Craft | Domain-touching |
| Slide 6 ("What year two looks like") | Q6 | Philanthropic Financial & Regulatory Marketing-Craft | Multi-domain |
| Closing slide ("Why trust us") | Q7 | Evidence & Brand Philanthropic Marketing-Craft | Multi-domain |
/give
| Paragraph target | Source Q | Domain lines triggered | Quorum |
|---|---|---|---|
| Lede paragraph ("Why LWD exists") | Q1 | Philanthropic Marketing-Craft | Domain-touching |
| "What we do" primary insert | Q3 | Philanthropic Financial & Regulatory Marketing-Craft | Multi-domain |
| "Before you give" callout (dignity floor made visible) | Q5 | Philanthropic Evidence & Brand Marketing-Craft | Multi-domain |
/give-sprint48-brief
| Paragraph target | Source Q | Domain lines triggered | Quorum |
|---|---|---|---|
| "What we say when we describe LWD" example paragraph | Q3 | Philanthropic Financial & Regulatory Marketing-Craft | Multi-domain |
| "What we intend for Y2" placeholder paragraph | Q6 | Philanthropic Financial & Regulatory Marketing-Craft | Multi-domain |
4. Compliance guardrails on every substitution
Every substitution the AI CMO proposes must survive the following checks before the review pack is assembled. If a check fails, the verdict is FIX-INLINE if the fix is mechanical, or RETURN-TO-SPECIALIST if the AI CMO cannot repair the substitution while preserving Lisa's voice. It is never a CLEAN that a human has to catch.
- Compliance markers survive. ABN, ACN, ACL 387398, mediation disclosure sentence, complaints link, privacy link. If a substitution accidentally removes any of these from a page, the substitution is rejected.
- Y1 numbers survive verbatim. 31 households · $1.82M consumer credit debt resolved · 23-day average case cycle · 52% average K10 reduction · $1,840 average cost per case · 87c per dollar reaching direct service. No paraphrase drift — if Lisa's voice-note rounds a number, the AI CMO restores the exact canonical figure.
- Outcome caveat attached where Y1 numbers appear. "These are Y1 outcomes for 31 households in a pilot period. They describe past outcomes only and are not a guarantee of individual future results." Substitutions that touch Y1 numbers must preserve the caveat.
- Dignity floor honoured. No beneficiaries are named or identified in any substitution. If Lisa's answer described a specific person, the identifiers have already been stripped by Corrina in the transcription pass — but the AI CMO checks again before assembling the review pack. Also: no hero-framing of carers, clinicians, or LWD staff; no "debt wipe" language; no urgency-appeal.
- Pre-DGR discipline. No tax-deductible messaging appears in any substitution until DGR is granted. If Lisa's voice-note references tax deductibility, the AI CMO holds that fragment in a "post-DGR" bucket and it does not enter the review pack until DGR status changes.
- Voice register preserved. Qualified rather than certain ("we expect" / "we intend" — not "we will" / "we guarantee"). Warm but non-appeal. Evidence-led. Quiet rather than urgent. Speaks about a specific person even when the person is anonymous. If the AI CMO's assembly drifts to charity-generic register, the verdict is
RETURN-TO-SPECIALIST. - Reversibility named. Every review pack states, in the reversibility field, exactly how the substitution unwinds if a human vetoes: "revert to previous paragraph text stored at
git-sha:X." No substitution is a one-way door.
5. Quorum forecast — how many review packs Sprint 55 produces
Counting the substitution rows across the seven page-cards above: ~30 substitutions total. Each substitution produces one review pack. So Sprint 55 lands ~30 review packs in the shared WhatsApp group.
| Quorum level | Approx. count | Signatories needed | Rationale |
|---|---|---|---|
| Domain-touching | ~7 | Owning human (Lisa) + one second-eye (Corrina by default; Laurence if Evidence & Brand touched) | Single-domain paragraph, no Y1 number touched, no forward-look, no financial claim. |
| Multi-domain | ~23 | Every owning human signs their line | Any paragraph touching Y1 numbers (Financial & Regulatory), forward-look (Financial & Regulatory), or the anchor-image / brand-voice work (Evidence & Brand) triggers a second line alongside Philanthropic. |
| Constitutional | 0 by design | All 4 humans | Sprint 55 is engineered to never touch the dignity floor. If a substitution accidentally trips constitutional (e.g. AI CMO's assembly produces an identifying detail), the verdict is RETURN-TO-SPECIALIST and Corrina rewrites the substitution before it reaches the group. |
Practically: Lisa signs her Philanthropic line ~30 times (once per substitution — and yes, this is deliberate: she reads the AI CMO's assembly of her own voice and confirms it before it ships). Corrina signs Marketing-Craft ~30 times. Carla signs Financial & Regulatory ~15 times (only where Y1 numbers, forward-look, or financial claims touch). Laurence signs Evidence & Brand ~10 times (anchor-image work, "why trust" paragraphs, brand-voice anchors).
Batching. ~30 review packs in one WhatsApp thread is too much. Corrina batches by page — one thread per donor page, with the 3–7 review packs for that page grouped inside it. Seven threads over ~48 hours is manageable. Lisa can work through one thread per sitting; each is ~10 minutes.
6. Unwind — how to revert a substitution if a human vetoes
If Lisa reads a review pack and says "no, that's not my voice" — or Laurence, Carla, or Corrina flag a problem — the substitution reverts to the previous paragraph text. The unwind is mechanical:
- Corrina posts "veto on pack #N" in the WhatsApp thread.
- The AI CMO holds the substitution out of the Sprint 55 commit.
- The other 29 substitutions ship as normal.
- The vetoed substitution is either abandoned (some paragraphs are best left as-is) or fed back to Lisa as "we couldn't hear your voice in this — would you re-answer Q{N} specifically for this paragraph?" — which is a tiny follow-on Sprint 55.1.
The unwind mechanic is exactly why substitutions are surgical rather than a rewrite: no substitution depends on any other, so vetoing one does not stall the rest. That is the whole point of the "30 small decisions, not one big one" architecture.
7. Change control
The map above is Sprint 55's contract. If a paragraph target changes (e.g. the AI CMO decides a Q3 answer is better placed in a different paragraph on /give than the one named here), that is a substitution amendment: single review pack, Marketing-Craft line (Corrina signs) + second-eye Laurence or Carla depending on which frameworks the paragraph touches. Adding a new target paragraph beyond the ~30 already mapped is a multi-domain amendment (Philanthropic + Marketing-Craft + Evidence & Brand). Adding a new source question beyond the seven in /lisa-voice-brief is a Sprint 55.1 rather than a Sprint 55 amendment — new questions require Lisa to answer them first, which is a fresh voice-capture ask.