Seven AI-drafted answers on the left. Space for Lisa's corrections on the right. Reduces the voice-capture cycle from "answer seven questions from scratch" to "read seven drafts, mark what's wrong, correct the register in the three or four places that don't sound like me." Prints cleanly on A4 for pen-and-paper markup; also copy-paste-friendly for WhatsApp text corrections.
What this is. The AI CMO has produced a persona-drafted version of each of the seven voice-capture answers — written to the tuning-fork register defined in /voice-brief (warm but non-appeal, evidence-led, qualified rather than certain, quiet rather than urgent). The drafts are not shipping anywhere. They exist for Lisa to react to, so that "editing" replaces "drafting from cold" as the cognitive task.
Why this shape. Editing is materially lighter cognitive work than drafting. If the drafts are 70–85% right in register, Lisa's job becomes catching the 15–30% that isn't her voice — which is fast, and which is exactly the correction the substitution machinery in /voice-map is built to consume.
What Lisa is NOT signing. This edit sheet is not the Philanthropic sign-off. It is the raw-material capture pass. The real Philanthropic line signs come later — once the AI CMO uses Lisa's corrections to produce ~30 paragraph substitutions and lands them in the shared WhatsApp group as review packs (per Sprint 54 machinery). Lisa signs each substitution then, not now.
Companion pages./voice-brief (the seven questions with context and anti-patterns) · /voice-map (the substitution index).
How to use this page (four options, pick whichever suits)
Print + pen-and-paper markup, then WhatsApp-photo it back. Print to A4 (Ctrl/Cmd + P; the print stylesheet strips navigation and makes the right column a clean 140mm dashed-border box per question). Mark up in ink. WhatsApp the photos to Corrina. This is often the fastest path.
Read the drafts here, type corrections into WhatsApp. One WhatsApp message per question: "Q3 — change 'four to seven active debts' to 'usually five or six.' Rest of Q3 fine." That's enough. Corrina and the AI CMO turn short deltas into full substitutions.
Voice-note the corrections. Read each draft here, then send a WhatsApp voice-note per question saying what to change. Preserves cadence; costs Lisa nothing but voice.
"This one is fine, ship it" is a valid answer per question. If a draft reads as Lisa's actual voice, say so — the AI CMO will use the draft as-is (still routed through the Sprint 54 gate, still signed on the Philanthropic line at review-pack time).
Time budget: ~10–15 min if editing (vs ~15–25 min if drafting from cold). No pressure to do it in one sitting — per-question is fine.
Two register warnings before you read the drafts. The AI added some things Lisa didn't explicitly say — a career sketch in Q1, dialogue framing in Q2, a specific media-request example in Q7, projection ranges in Q6. These are inventions in service of the register, not facts about Lisa. If any of them is wrong (or overreaches into biography Lisa doesn't want in donor copy), strike it. The right column is where those strikes go.
New financial figures beyond the Y1 canonical set appear in Q3 ($59k per-household debt average, 4–7 active debts), Q6 (100–120 Y2 households, 15–25 referrer relationships, ≥80c per-dollar floor), and Q7 (salary sector-benchmark). Per the anti-pattern list these need Carla's Financial & Regulatory sign-off before landing on any donor page. Either Lisa validates them here (which means they're real numbers, not AI inventions), or they get stripped from the substitutions.
Biography invention The AI added a career backstory ("twenty-two years across three careers... institutional philanthropy on the giving side"). If wrong or overreaches, strike.
Near-identifier "a conflict-of-interest recusal from a foundation grant round in 2023" — is this true, and is it safe to say publicly? If it identifies the foundation or the person, either strip the year, strip the specifics, or replace with "some years ago."
Register "shape of hole in it" is a strong metaphor. Lisa may or may not talk like that.
AI draft (persona pass)
I came to charity work the long way around. Twenty-two years across three careers before this — the last one in institutional philanthropy on the giving side, sitting on the other side of the table from organisations asking me for money. So I've read a very large number of grant briefings, and I've been trained to spot the difference between a charity that has an answer and a charity that has a story.
The reason LWD exists for me personally is a small one. I sat with a woman during a conflict-of-interest recusal from a foundation grant round — she wasn't the applicant, she was a friend of a friend, and she wanted five minutes of unofficial advice about what to do with the debt letters that were arriving during her husband's cancer treatment. And what I realised in that five minutes is that the entire Australian charity sector had a shape of hole in it that she was falling through. There were hardship charities that gave food. There were financial counsellors who gave advice. There were credit ombudsmen who resolved disputes after the fact. Nobody was doing the thing she actually needed, which was: sit between me and my creditors while I am in this situation, and negotiate an outcome I can meet, using an authority I don't have but you do.
That authority is Australian Credit Licence 387398. That five-minute conversation was the moment I understood what LWD would need to become — not another hardship charity, not another advocacy body, but a mediation service that holds a licence and uses it. Laurence had already been working on the model for eighteen months by then; I brought the philanthropic side. That is why I am here, and why this rather than something adjacent.
Lisa — corrections here (or ship as-is)
Common correction shapes:
"Career sketch is wrong — my background is [X]"
"Strike the 'twenty-two years' sentence — too specific / not relevant"
"Change 'friend of a friend' to [Lisa's actual phrasing]"
"Strip the year / the foundation reference — not for public copy"
"Rest is fine, ship it"
If pressed for time on Q1, at minimum check
Whether the biographical sketch is anywhere near true. Everything else can be corrected downstream by the AI CMO; a wrong biography can't.
AI CMO consumption format · Q1Q1 draft status: [ship-as-is | edit-below | reject-redraft]
Lisa's corrections (verbatim):
<paste here or leave blank if ship-as-is>
Biographical facts to preserve if edited:
- [true?] LWD is Lisa's first charity role (implied by "new to charity work")
- [true?] Prior career included time on the philanthropic-giver side
- [true?] Meeting with the described household predated LWD's founding
- [true?] Laurence's model predated Lisa's involvement by ~18 months
Domain lines to trigger for the resulting substitutions:
Philanthropic (Lisa) + Marketing-Craft (Corrina) + Evidence & Brand (Laurence, second-eye on founder-story consistency)
2
What is the moment you'd want a foundation trustee to picture?
Triangulation risk "One of the eastern-seaboard capitals" + gender pairing + specific medical detail (extended treatment plan by four cycles) — in a small charity sector these could triangulate. Strong candidate for stripping the eastern-seaboard reference and softening the treatment specifics.
Invention "three unopened envelopes stacked between two cold cups of tea" — specific enough that it needs to be true, or clearly acknowledged as composite. Lisa should confirm.
Register "the letters had become the thing consuming the household" — strong but potentially heavy. Lisa's actual sentence may be quieter.
AI draft (persona pass)
A dining table in a suburb I won't name, in one of the eastern-seaboard capitals. Three unopened envelopes stacked between two cold cups of tea. The person opposite me had just come back from a hospital appointment where the treatment plan had been extended by four cycles, and she was trying to work out — out loud, not asking me, just working it out — whether to open the envelopes before or after her husband got home from the same appointment.
I did not open the envelopes. She did not open the envelopes. That was not what the conversation needed. What the conversation needed was for someone to notice that the letters had become the thing consuming the household — not the illness, not the treatment, not the prognosis. The letters. Three unopened envelopes were doing more damage than the disease at that moment, because the disease had a medical team and the letters had nobody.
That is the moment. Not the diagnosis, not the treatment, not the recovery, not the loss. The moment where the debt correspondence becomes the household's fourth crisis stacked on top of the first three, and nobody is standing in it. LWD stands in that moment. Everything else we do — the ACL, the mediation, the audited outcomes, the referrer network — exists to let us stand in that moment competently.
Lisa — corrections here (or ship as-is)
This one carries the most dignity-floor risk — three specific asks:
Is the moment true, or should we mark it "composite"?
Soften "extended by four cycles" to something less identifying?
If pressed for time on Q2, at minimum check
That the anchor image doesn't identify anyone. If it does, either strip identifiers or replace with an explicit composite disclosure ("a composite of moments I've been in") — the trustee reader accepts either honestly.
What does the mediation function actually do, in your words?
Answers into every donor page's "what LWD does" paragraph · sits adjacent to the ACL 387398 mediation-disclosure sentence · primary insert on /give
Register warnings on this draft
New figures "around fifty-nine thousand dollars per household" and "four to seven active debts" — NEITHER is in the Y1 canonical six. Lisa needs to confirm these are real pilot figures. If yes, they become new figures Carla signs into permanence. If no, strip both sentences.
Compliance-adjacent "sometimes it is a full waiver where the creditor's own hardship framework applies" — is this actually how any Y1 case resolved, or is it an idealised description? If never, strip.
Register "extinguish the impossible version of the debt" — strong metaphor. Fits Lisa's brief but is a candidate for Laurence's Evidence & Brand line to check that the metaphor is on-brand and not accidentally misleading under RG 96.
AI draft (persona pass)
Mediation is the negotiation the household cannot do for themselves, done under a licence that gives the negotiation force.
Here is the shape of it. A household comes to us — usually referred by a hospital social worker or a palliative-care team, occasionally by a financial counsellor. They arrive with, on average, four to seven active debts across a mix of credit cards, buy-now-pay-later, energy, and unsecured personal loans. Total varies; the pilot average was around fifty-nine thousand dollars per household. Every one of those creditors has a legally sanctioned process for what happens next if the household does nothing — and that process ends, eventually, in default judgements, wage garnishment, or bankruptcy.
What we do is: we contact each creditor under ACL 387398, we present the household's medical circumstances and financial position with the household's written consent, and we negotiate a settlement. Sometimes that settlement is a reduced balance paid over a period the household can meet. Sometimes it is a hardship variation that pauses collection until treatment concludes. Sometimes it is a full waiver where the creditor's own hardship framework applies. What lands on the table is a set of settlements that, together, the household can actually meet from their actual income.
The debt does not vanish. LWD does not extinguish debt. What we extinguish is the impossible version of the debt — the version where twelve creditors are all pursuing full recovery simultaneously against a household in medical crisis. What replaces it is a version the household can pay, on a timeline that fits their life, without losing the house or the car or the dignity.
The Year One audited outcome was that we did this for thirty-one households and the total debt resolved across those households was $1.82M. That is not $1.82M in donations. That is $1.82M in creditor claims that were pursued against households in serious illness, that we mediated into settlements the households could meet. The distinction matters more than any other single sentence I could say about LWD.
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.
Lisa — corrections here (or ship as-is)
Highest-value edits:
Are "$59k per household" and "4–7 active debts" the real Y1 averages? (If Lisa doesn't know: mark "check with Carla" — do not ship.)
Is the three-way settlement typology (reduced-balance / hardship-variation / full-waiver) accurate to Y1?
Any dial-back on the "impossible version of the debt" line?
Rest fine as-is?
If pressed for time on Q3, at minimum check
The two new figures ($59k, 4–7 debts). This paragraph feeds every donor page's "what we do" section, so getting the figures right (or explicitly stripping them) matters more than the metaphors.
AI CMO consumption format · Q3Q3 draft status: [ship-as-is | edit-below | reject-redraft]
New-figure validation (Carla must sign before any of these ship):
- "$59k per household average" : [confirmed | wrong-figure-is-X | strip]
- "4-7 active debts per household" : [confirmed | wrong-figure-is-X | strip]
- "reduced-balance / hardship / waiver" : [accurate | partial-accurate | strip]
Lisa's corrections (verbatim):
<paste here>
Domain lines to trigger:
Philanthropic (Lisa) + Financial & Regulatory (Carla, on the figures) + Marketing-Craft (Corrina)
Quorum: Multi-domain
4
Why 31 households and $1.82M — and not something bigger?
Only potential edit is stylistic — whether Lisa's actual voice frames "the discipline of thirty-one" that way, or differently.
AI draft (persona pass)
Because the number needed to be small enough to measure properly, and thirty-one was the number at which the measurements began to hold their shape.
Let me explain what I mean by that, because "we deliberately stayed small" sounds like a defence, and it isn't one. It is a design choice.
If we had taken on eighty households in Year One, four things would have degraded. First: the K10 psychological-distress instrument — which is our primary well-being measure — needs an intake reading and a mid-cycle reading and an exit reading, and every one of those readings takes a trained conversation. At eighty households we could not have run those conversations at the quality that lets the fifty-two percent reduction figure mean anything. It would have become a checkbox instead of a measurement. Second: the twenty-three-day average case cycle depends on hand-on-the-file oversight from a very small team. At eighty we would have queued cases, and case-cycle time would have doubled, and the twenty-three-day figure would have stopped describing what we actually do. Third: the eighty-seven cents per dollar reaching direct service depends on a lean overhead structure that only works at pilot scale. Scaling requires investing in that overhead — the very investment we are seeking now — and doing that investment before we had the audited evidence to justify it would have been the wrong sequence. Fourth: the eighteen-hundred-and-forty-dollar cost per case is a real number because we could interrogate every dollar. At eighty households, the accounting becomes averaged rather than known.
So Year One was small on purpose, and it was small at the number where the model held together honestly. Thirty-one is the number at which every one of the six audited figures can be defended, dollar by dollar, case by case. If a foundation trustee wants to spend a day going through our workings, we can walk them through all thirty-one files and the numbers will hold. That is what we were protecting for.
Year Two exists to scale the model. But scaling honestly requires that Year One proved the model, and proving the model requires that Year One was small. That is the discipline of thirty-one.
Lisa — corrections here (or ship as-is)
This one is likely ship-as-is or light-edit:
Does "the discipline of thirty-one" as a closing line sound like Lisa?
Is the four-part degradation argument in the right order for Lisa?
Any of the four points overreach (e.g. are we sure at "eighty" queueing doubles the case cycle, or is that a plausible-but-invented figure)?
If pressed for time on Q4, "ship as-is" is a strong default
Q4 is the draft that's closest to clean. Even a rushed edit shouldn't touch much.
AI CMO consumption format · Q4Q4 draft status: [ship-as-is | edit-below | reject-redraft]
Y1 figures used (must remain verbatim):
- 31 households
- K10 52% reduction
- 23-day case cycle
- 87c per dollar direct service
- $1,840 per case
All present. Outcome caveat inherited from Q3 substitution.
Lisa's corrections (verbatim):
<paste here>
Domain lines to trigger:
Philanthropic (Lisa) + Financial & Regulatory (Carla, on the four-part argument) + Marketing-Craft (Corrina)
Quorum: Multi-domain
5
What is the wrong reason to give to LWD?
Answers into /donor-stewardship-plan "who this is for" · /give "before you give" callout · /founding-donor-deck slide 5
Register warnings on this draft
Comparative charity language "There are excellent Australian charities that offer that giving experience honourably." Some governance frameworks discourage comparative charity marketing. Needs Laurence's Evidence & Brand second-eye. Recommended: keep the sentiment, soften "offer that giving experience" if it reads as steering donors elsewhere.
RG 96 adjacent The paragraph explicitly names "just one more dollar could save a family from bankruptcy today" as a banned pattern. That is on-brand but it prints the anti-pattern — Corrina should decide whether the anti-pattern appears in donor copy verbatim or is described obliquely.
Directness "There will be no [X]" × several is strong. Fits Lisa's brief but is the strongest tonal choice of the seven — Lisa may soften.
AI draft (persona pass)
If you are looking for a rescue-narrative charity — the kind that publishes photographs of the people it has helped, that runs "meet the family you saved" campaigns, that tells you at the end of the tax year what your particular donation specifically achieved for a specific named recipient — we are the wrong charity for that giving experience. And I want to say this cleanly, because the temptation to soften it is real, and softening it would be a lie.
We will not put a beneficiary's face on the front of a document. We will not name a household, ever. We will not publish quotes from people we have helped, even with consent — and even if we had consent, we would decline. We do not have a "success story" section on the website and we will not build one. The dignity floor we hold is not a marketing choice; it is a governance one, and it is not negotiable.
We are also not an urgency charity. There will be no "just one more dollar could save a family from bankruptcy today" appeals. There will be no countdown clocks on the give page. There will be no matched-giving campaigns timed to a hospital diagnosis calendar. Australian Securities and Investments Commission Regulatory Guide 96 has clear conduct rules on credit-adjacent language, and the register those appeals require would breach them even if we were willing — which we are not.
If the giving experience you are seeking is one where you can point at a named person and say "I helped that one" — there are excellent Australian charities that offer that giving experience honourably, and I would rather steer you to one of them than compromise how we operate to fit you. Our donor is a different donor. Our donor is one who wants dignified evidence over dignified stories, audited numbers over anecdotes, and the professional quiet that comes with treating philanthropy as an investment in a working model rather than as an emotional purchase.
Saying that clearly at the front door is not us being difficult. It is us respecting your time and ours. The donors who stay after reading that paragraph are the donors we can serve well for the long term.
Lisa — corrections here (or ship as-is)
This is the paragraph that shapes donor self-selection. Worth extra time:
Keep the comparative sentence, or soften it?
Print the banned appeal wording ("just one more dollar...") in donor copy, or describe obliquely?
Anything in this paragraph that would embarrass Lisa if a specific existing prospective donor read it and recognised themselves?
"An emotional purchase" — too sharp?
If pressed for time on Q5, at minimum check
The comparative-charity sentence (Laurence flag) and the printed anti-pattern quote. Everything else can hold.
Answers into /foundations-grant-briefing "what year two looks like" · /founding-donor-deck slide 6 · /donor-stewardship-plan forward-look
Register warnings on this draft
New projections "100–120 households in Y2," "15–25 referrer relationships," "≥80c per-dollar Y3 floor." All three are AI projections. Need Carla's Financial & Regulatory sign-off before shipping to any donor page.
New figure "Q3 decision on second city" — is there actually a Q3 decision milestone, or is this AI framing? Lisa/Laurence confirm.
Register All hedges present ("we expect," "we intend," no "we will"). Good. But the volume of projection may still feel over-committing — Lisa may want to strip a range or two.
AI draft (persona pass)
We expect Year Two to take LWD from thirty-one households to somewhere between one hundred and one hundred and twenty. That range is deliberate. The lower end is what we can reach with the current team and one additional case-worker hire. The upper end assumes the DGR grant lands within the current window and the philanthropic pipeline responds to that as forecast. Between those two scenarios sits the honest range of what Year Two looks like if things work.
What we expect to hold, across that range, are the three model-describing figures. The fifty-two percent average K10 reduction should hold, because it describes the intervention rather than the scale. The twenty-three-day average case cycle should hold, provided the new hire is in place by the end of Q2 — otherwise it drifts, and we would rather report the drift honestly than defend a number that no longer describes what we do. The eighty-seven cents per dollar reaching direct service is the one I hold most carefully. It should hold at one hundred households; it will inevitably degrade at two hundred, three hundred, five hundred as we invest in the infrastructure that lets us scale. We are already thinking about what our floor is. We intend the floor to be no lower than eighty cents per dollar direct-service at steady-state Year Three, and if we cannot hold that floor honestly we will slow the growth curve until we can.
Geographic reach in Year Two moves from the current single-city anchor to a second-city trial. The decision on which second city is Q3, and it depends on two things: which hospital referrer network has the capacity to absorb us, and which financial counselling sector we can partner with rather than compete against. That decision is not made yet, and I would rather tell a trustee "we don't know yet, and here is the framework for how we will decide" than name a city we might not choose.
The referrer network is where we grow only as fast as we can be trusted. Hospital social workers are the hardest audience we have, because they refer households and they are professionally accountable for who they refer to. They will refer to LWD once. If we do the work well, they refer again, and the referrer network grows organically. If we do the work poorly on a single household, that social worker's team stops referring for years. So the referrer network is not a growth lever we can push; it is a trust structure we tend. Year Two grows the number of hospital teams referring to us from the current small number to somewhere between fifteen and twenty-five active referrer relationships. Anything faster than that would be dishonest growth.
And I want to be clear on one thing about all of these numbers: they are what we expect and what we intend. They are not what we guarantee. Charity is not the sector for guarantees, and any organisation that offers you one has told you something else about itself in the offering.
Lisa — corrections here (or ship as-is)
Every projection needs a yes/no from Lisa (or "check with Laurence"):
100–120 households Y2 range: right, wrong-figure-is-X, or "check"?
≥80c Y3 floor commitment: right, wrong, or too specific?
15–25 referrer relationships: right, wrong, or "check with Laurence"?
"Q3 second-city decision" — real milestone or invention?
"Two hundred, three hundred, five hundred" implied trajectory: keep or strip?
If pressed for time on Q6, at minimum check
That the three named projections (households, referrers, floor) are ones LWD wants on the record. Everything on this page ends up in front of foundation trustees who read forward-looks carefully.
Specific incident "A television current-affairs programme approached us in April" + "three journalists' names are on file" — is this true? If invented (even as a plausibility example), strip. If true, is April this year or an earlier year? Is naming journalists on file real or aspirational governance?
New figure "Neither of us is drawing a salary above the sector benchmark" — introduces a salary claim. Needs Carla's Financial & Regulatory sign to appear in donor copy. Alternative: strip or generalise ("salaries are within sector norms and disclosed in the annual report").
Length This is deliberately the longest answer — the anchor. Every donor page it lands on may want a shorter version. Lisa may write a "short form" while she's at it: the two or three sentences to lead with when a trustee only reads the first paragraph.
AI draft (persona pass)
Because we have done thirty-one, and we have the receipts. Every one of those thirty-one files is auditable. The K10 readings are in the case-management system with dates and case-worker attribution. The debt-resolution figures are in the accounting system, reconciled against creditor correspondence held on file. The cost-per-case figure is Carla's audit, and Carla is our external CFO from COSAI, and she will show you exactly where every dollar of the eighteen-forty went and exactly where the thirteen cents that did not reach direct service went.
Because we hold Australian Credit Licence 387398, and we have operated inside it for the entire pilot. That licence carries a set of NCCP-compliant conduct rules that are enforced by regulators, not by us, and every case we run is compliant with them by design. There is no cases-off-the-books, no informal mediation, no going outside the licence. If we ever needed to, we would close the pilot rather than do that.
Because we have said no to things we would have benefited from saying yes to. A television current-affairs programme approached us in April looking for a family to feature on camera during a treatment cycle. It would have grown our donor pipeline significantly and rapidly. We declined, because putting a beneficiary's face on the front of a national broadcast would have breached the dignity floor, and the dignity floor is not for sale. We turned down two smaller similar approaches earlier this year for the same reason. If a trustee wants to verify that: three journalists' names are on file with our records officer, and each of them will confirm we declined and why.
Because Laurence is the Chief Executive Officer of record and I am the accountable person for how philanthropic money is spent, and neither of us is drawing a salary above the sector benchmark for our roles, and our salaries are in the annual report line-by-line. There is no shadow governance here. What you see on the org chart is the operating structure, not the presentation structure.
Because we have external contractors — Carla at COSAI on the financial side, Corrina at Your Digital Team on the marketing side — and they hold their contracted domains with full weight, but neither of them clears constitutional questions alone. Those questions come to Laurence and me. The org has a shape, and the shape holds under pressure.
And because we would rather grow honestly than grow quickly. If Year Two proves that the model does not scale the way we expect it to, we will report that outcome honestly and slow the curve, not paper over it and press on. Year One's small number was the first proof of that discipline. Year Two, whatever it looks like, will be the second. The right question for a trustee to ask is not "will you succeed" — every charity says yes to that. The right question is "will you tell us honestly if you don't." That is the trust I am asking for. It is the trust I am accountable for. And it is the trust the audit trail is designed to make verifiable, quarter by quarter, so you do not have to take my word for any of it.
Lisa — corrections here (or ship as-is)
This is the anchor answer. Three specific asks:
Is the "television programme in April" incident true? If not, strip that paragraph and replace with a real example (or with a general "we've declined media approaches that required beneficiary-facing content"). Lisa or Laurence.
Salary paragraph — keep the specific "sector benchmark" claim, or generalise? (Carla's call as well.)
Optional: write a two-sentence short-form version. The full answer runs long; every donor page needs the choice to lead with the short version and link to the full.
If pressed for time on Q7, at minimum check
The two factual claims (media incident, salary framing). This is the paragraph a trustee will remember; the two invented specifics are the two that can most damage credibility if they're wrong.
Three things worth flagging before Corrina couriers Lisa's corrections into the WhatsApp group as Sprint 55.1 raw material. These apply across the seven answers rather than per-question.
New financial figures. Every $ figure introduced beyond the Y1 canonical six (31 · $1.82M · 23-day · 52% · $1,840 · 87c) needs Carla's Financial & Regulatory sign-off before appearing in donor copy. That list, from the drafts above, is: $59k per-household debt average (Q3), 4–7 active debts (Q3), 100–120 Y2 households (Q6), 15–25 referrer relationships (Q6), ≥80c/$ Y3 floor (Q6), sector-benchmark salary claim (Q7). Corrina should route these figures to Carla in a single Financial & Regulatory batch before the AI CMO composes any substitution using them.
Dignity-floor risks introduced by the drafts. Two identifier risks: the "eastern-seaboard capitals" phrasing in Q2, and the "April this year TV programme" specifics in Q7. Both are stripping candidates. Corrina should strip both before the AI CMO composes any substitution — even if Lisa doesn't explicitly flag them, the default is remove.
Register drift candidates. Three phrases that read as heavier than Lisa's brief register: "shape of hole in it" (Q1), "consuming the household" (Q2), "extinguish the impossible version of the debt" (Q3), "emotional purchase" (Q5). Any of these Lisa signals a preference against, the AI CMO substitutes with plainer language. Any Lisa signals a preference for, the AI CMO keeps — but only for the specific paragraphs the substitution index puts them in.
How the AI CMO consumes this once Lisa is done
Once Lisa returns the seven corrections (via any of the four correction paths in the instruction band above), Corrina assembles a single AI CMO session-launch paste that folds in each answer's status and Lisa's corrections. The paste extends the standard Sprint 55 session-launch paste from /voice-brief.
Sprint 55.1 AI CMO session-launch paste (paste after the 4 standard AI CMO blocks + the Sprint 55 voice-insertion paste)Sprint 55.1 refinement task: consume Lisa's corrections from the edit sheet and adjust the substitution proposals accordingly.
Per-question consumption (each block below is prefilled from the edit sheet):
Q1:
status: [from edit sheet]
corrections: [from edit sheet]
biographical facts confirmed: [from edit sheet]
Q2:
status: [from edit sheet]
composite disclosure: [from edit sheet]
identifiers stripped: [from edit sheet]
corrections: [from edit sheet]
Q3:
status: [from edit sheet]
new-figure validation from Carla: [pending Carla sign | confirmed | stripped]
corrections: [from edit sheet]
Q4:
status: [from edit sheet]
corrections: [from edit sheet]
Q5:
status: [from edit sheet]
comparative sentence handling: [from edit sheet]
printed anti-pattern quote: [from edit sheet]
corrections: [from edit sheet]
Q6:
status: [from edit sheet]
projection validations from Carla + Laurence: [pending | confirmed | stripped]
corrections: [from edit sheet]
Q7:
status: [from edit sheet]
factual claims validated: [from edit sheet]
short-form version: [from edit sheet]
corrections: [from edit sheet]
Instruction:
For each of the ~30 paragraph substitutions in /donor-voice-map, produce a review pack containing:
- Current paragraph (before)
- Proposed paragraph (after) - assembled from the corrected Lisa answer(s), NOT the AI draft
- Source Q number
- Confidence note - specifically whether this paragraph uses any "pending Carla sign" figures (if yes, verdict = CLEAN+FLAG until Carla signs)
- Domain lines triggered
- Quorum level
- Reversibility note
Any paragraph that would use a Q3/Q6/Q7 new-figure without Carla's sign, or a Q2/Q7 identifier without Lisa's explicit "keep," gets held in the pack queue with verdict RETURN-TO-SPECIALIST.
Compliance markers must survive every substitution: ABN, ACN, ACL 387398, mediation disclosure sentence, "complaints" and "privacy" links intact.
No beneficiaries are named or identified in any resulting substitution.
Expected turnaround from Lisa's corrections landing to review packs in WhatsApp. ~2 hours of AI CMO time (unattended, running through the substitution machinery) + ~30 min of Corrina courier time (batching by page and posting into WhatsApp) = same-day flip from Lisa's answers to first review packs in the group. Lisa's own sign-offs (Philanthropic line, ~30 packs) then spread over the following ~48 hours, one WhatsApp thread per donor page.