The Legacy Channel — eldercare GTM (Living Voice Track C, deep dive)

Status: strategy / not built. Authored 2026-07-08. This is the detailed go-to-market for Living Voice Track C (the enduring voice, sold into eldercare), and it revises the Track C strawman in LIVING_VOICE_ROADMAP.md: the roadmap assumed a B2B per-facility license; this doc argues the real wedge is the adult child at the moment of placing a parent in care — a motivated, paying, guilt-carrying buyer who brings the product into the facility. Facility = distribution channel, not the buyer. Grounded in a cited market brief + a full inventory of the shipped Lovio infra (2026-07-08). Not a pull-forward from Workstream A / Odessa.

1. The reframe (why the wedge flips the whole model)

Track C originally read as classic B2B: sign facilities, license per bed. That's a ~70–150-day sales cycle against free incumbents ("This is me," Music & Memory) with the Executive Director as a budget gatekeeper — slow, and you're a line item competing for enrichment dollars.

The better shape: the buyer is the adult child, at placement. They feel acute, near-universal guilt; they are actively hunting for a way to "stay connected"; they already write $5,900–$6,160/mo checks for the care itself; and — critically — they are the legally clean contracting party (keeps us out of HIPAA scope, see §7). The facility doesn't buy; it refers — because the product makes families happier and operationalizes the reminiscence programming activity directors already believe in. This turns a cold enterprise sale into a warm, emotional, consumer purchase with the facility as a trusted amplifier.

2. The buyer & the moment

Placement guilt is documented and near-universal, and adult children carry it worse than spouses.

• 50% of caregivers feel "somewhat guilty" about placement, >13% "extremely guilty" (AARP). Adult children vs. spouses: 37.5% vs. 26.7% placement guilt (dementia-caregiver RCT sub-study).

• It's framed clinically as a normal, long-lasting "no-win" state, layered on anticipatory grief / ambiguous loss ("physical presence, psychological absence").

Three high-intent windows — concentrate on the first two:

1. Pre-move-in / decision period — peak stress, active research, "getting affairs in order."
2. First ~90 days after move-in — the Relocation Stress Syndrome / "transfer trauma" window (a recognized nursing diagnosis; ~30% of transferred residents meet criteria). Guilt is freshest; the child is looking for a way to reconnect. This is the acquisition trigger.
3. Decline window — urgency to "capture the voice before it's gone." Real, but consent + recording quality degrade here (see §7) — a reason to pull capture earlier, not lean on this window.

Scale of the motivated buyer: 7.2M Americans have Alzheimer's in 2025 (→~13M by 2050); ~13M provide unpaid dementia care (>⅓ are daughters); dementia caregivers report severe+ stress at 37.2% vs 21.4%. And StoryWorth has recorded >35M stories / >1M printed books since 2013 — proof that families already pay to capture an elder's life story.

3. The strategic gift (three urgencies, one moment)

The single most important finding: the moment that maximizes buyer intent, clinical validity, AND legal cleanliness is the same moment — early-stage / at-or-before placement.

• Intent: guilt is acute at placement.
• Clinical: reminiscence/life-review evidence is strongest, and recordings are richer, while the parent is still verbal.
• Legal: the parent can give living, informed, voice-specific consent while they still have capacity — the legal safe harbor (see §7).

Everything points at the same window. Design the entire funnel to hit "we just moved Mom in" (and ideally "we're about to").

4. The human journey

A daughter moves her father into memory care. Two weeks in, guilt-ridden and grieving the version of him that's fading, she gets a gentle referral (from the facility's welcome packet, the activity director, or a targeted ad): "Capture Dad's voice and stories — while there's still time, in his own words."

She sets up an account (hers), and starts a facilitated capture for her father: she sits with him with a phone, and the app walks them through guided reminiscence prompts — "Tell me about the day I was born," "What was your father like?" He talks; it records. His voice is cloned from the clean samples. Over several short sessions (gentler than one long take), a life-story archive builds.

She composes capsules for the whole family — one for each grandchild's 18th birthday, one for her sister, one for herself. Some are delivered now; some are sealed to unlock after he's gone. When that day comes, the family receives him — his stories, in his voice — on their terms, not pushed.

Every hard requirement in the product falls out of this story: facilitation (she operates, he is the subject), guided prompts, multi-recipient family delivery, after-death delivery, and consent that survives him.

5. The competitive whitespace

The unoccupied square: cloned voice + delayed/after-death delivery + care-facility distribution, in one offering. Only Eternos is genuinely in the voice+after-death lane, and it has no facility channel. LifeBio owns the facility channel but has no consumer voice-clone or delivery product.

The graveyard is the warning. HereAfter (shutting down) and StoryFile (Ch. 11) prove server-dependent "talk to the dead" playback is a durability/trust liability. Our product must credibly promise the recordings outlive the company — downloadable master files, family ownership, ideally an escrow/export guarantee. This is also a differentiator: "you own the files forever," not "trust our servers."

6. The offer & pricing (family-paid)

The deliverable is not a subscription-to-talk-to-the-dead (that model is dying). It's a permanent, owned family archive: the voice, the guided life-story recordings, and scheduled capsules — with the raw audio downloadable so it can never be orphaned.

Pricing lean (validate; anchored between StoryWorth $199 and Eternos $25/mo):

• One-time "capture" package (voice clone + N guided sessions + the family archive + downloadable masters) — a one-time fee families understand as a keepsake, not a SaaS bet. Anchor ~$149–299.
• Optional low monthly for ongoing capsule scheduling / new recordings / hosted playback — kept cheap because the value is in the owned archive, not the hosting.
• Facility never pays in the wedge model (see §8). A later enterprise tier is possible once the family motion proves out.

7. The hard parts, head-on (the ethics + legal wall)

This is the make-or-break, and it's a legal architecture problem, not a UX checkbox. Get counsel before building — the points below are from law-firm/estate-planning secondary sources.

a) The current consent model forbids this flow. The shipped Lovio consent is hard-coded first-person ("I consent to HiveJournal cloning my voice… will never narrate anyone else's content"), scoped auth.uid() = user_id. The child-facilitates-parent case has a facilitator (child) operating and a subject (parent) consenting — a distinction that does not exist in code today. Load-bearing build: a subject-consents / facilitator-operates consent model — the parent records a voice-specific consent statement (the proof artifact), the child operates the account. This is gap #1 and everything depends on it. Designed in detail in LEGACY_CHANNEL_CONSENT_MODEL.md (schema, the v3 read-aloud statement, capacity/surrogate workflow, heir-durability, and the questions for counsel).

b) Capacity is impaired in 44–69% of nursing-home residents, and it fluctuates. Dementia ≠ incapacity, but you cannot assume it. Two consequences: (1) capture consent while the parent still has capacity (early-stage / pre-placement — which is also the highest-intent, highest-quality-recording moment, per §3); (2) build a capacity + surrogate-consent workflow (guardian/POA/healthcare-proxy co-sign) for the diminished-capacity case, treated as first-class, not an edge case.

c) Voice/likeness rights flow through the estate, not POA. A financial POA and a healthcare proxy generally do not convey likeness/publicity rights — those pass through the estate/heirs. TN's ELVIS Act (2024) explicitly protects voice against AI cloning; CA AB 1836 (Jan 2025) bars unauthorized digital replicas of the deceased. So consent must be living, informed, voice-specific, and heir-durable — obtained from the person while alive, structured to bind their estate. A generic POA is not enough.

d) Keep the family as the contracting party to stay out of HIPAA. A consumer memory/legacy app with no BAA, not operating on a provider's behalf, falls outside HIPAA. The moment a facility or health plan pays and data flows on their behalf, you may pull yourself into HIPAA/BAA scope. This is an independent reason the wedge (family pays) is the right model — it's legally lighter, not just commercially warmer. (State consumer-health-data laws, e.g. WA's My Health My Data, can still reach it — design for sensitive-PII handling regardless.)

e) Surrogate consent sits in the elder-exploitation risk zone. Taking a surrogate's consent to record/clone a vulnerable adult is adjacent to recognized POA-abuse patterns. Mitigate with: the subject's own recorded consent wherever capacity allows, a documented capacity/surrogate workflow, revocation as a first-class action, and a human dignity review of prompts + framing.

f) "Is this exploitative in grief?" The dignity posture (inherited from Living Voice + Lighthouse): read-only by default — we narrate what they actually said; we do not generate new speech in a dead parent's mouth (the bright line). Framing is always "a recording of [name]", never resurrection. Delivery is on the recipient's terms, never a push. No growth-hack language anywhere near grief copy.

8. The facility relationship (channel, not buyer)

• Champion: the Activity / Life-Enrichment Director. NAAP defines the role as the resident's "connector"; they own reminiscence/life-story programming (where Music & Memory and "This is me" already live). They want this.
• Gatekeeper: the Executive Director (budget + resident access + family relations). But in the wedge model we're not asking for budget — we're offering a family-satisfaction win at zero cost to the facility, which is a far easier "yes."
• Bridge: the social worker (mandated family communication channel).
• Poor primary channel: hospice — the chaplain is a values-aligned ally, but the ~17-day median stay is too short. Concentrate at placement, not end-of-life.
• The facility's justification to participate is the reminiscence/life-review evidence, not our voice-clone tech: Cochrane finds reminiscence therapy's quality-of-life benefit concentrates in care-home settings (SMD 0.46); life-review therapy shows a large acute effect on late-life depression (g=1.41, though durability is uncertain). Frame institutional value as person-centered care + family satisfaction, and slot alongside proven precedents. (Do not overclaim the familiar-voice/Simulated-Presence evidence — it's real-but-very-low-quality; call it "promising," not proven.)

9. What we reuse vs. what's new

Reusable as-is (shipped Lovio infra):

• Voice clone create/retire + Quick/Studio tiers — lovio-voice.ts (createVoiceClone takes samples[]; many files = richer Studio clone — ideal for multi-session capture).
• Consent-recording-as-proof + consent_version plumbing (migration 458) — the mechanism is reusable even though the scope isn't.
• Capsule seal → render → stitch → snapshot, incl. long-script chunking — lovio-seal.ts.
• Date-triggered delayed delivery cron + idempotent claim-first send — lovio-delivery.ts.
• Token-based no-auth recipient unlock page + <PresenceViewer> / <SitWith> playback — the "sit with them" moment is already built (Track B, now hands-free via the shipped audio-first player).
• Per-entry voice render + cache, notebook-bound capsules, quarterly reminders.

Must-build (the eldercare-specific gaps, in dependency order):

1. Subject-consents / facilitator-operates consent model — §7a. The keystone; the current bright line forbids the flow without it.
2. Facilitated / assisted capture mode — operator↔subject separation (child's account manages the parent's clone + capsules).
3. Death-triggered delivery — the 'death' enum value exists but has no handler (seal rejects it; crons filter to 'date' only). Needs a dead-man's-switch / legacy-contact verification. This is the core eldercare requirement and is entirely absent.
4. Custody that outlives the owner — today capsules are ON DELETE CASCADE to the owner's account (they're destroyed if the owner leaves) — the opposite of a legacy product. Needs recipient accounts / claim flow / estate handoff + the "downloadable masters" durability promise (§5).
5. Multi-recipient / whole-family delivery — today a capsule has one recipient; a family needs N. Needs a recipient-group / fan-out model.
6. Guided reminiscence prompt library — zero exists; capture is free-form journal entries. Needs a curated prompt set driving the capture sessions.
7. Elder-friendly / facilitated UX — the current flow is dense + jargon-heavy; acceptable only because the child operates it, but the facilitated session UI (large targets, forgiving retries, session-based) is new.

10. Phasing

• C0 — one-family pilot (no facility). Build gaps #1, #2, #6 (consent model + facilitated capture + prompts) and run it with a real family capturing a real parent. Learn the consent + capacity workflow and the capture UX with the highest-stakes users before any facility is involved. Cheapest path to the core learning.
• C1 — after-death delivery + custody. Build gaps #3, #4, #5 (death trigger, custody-outlives-owner, family fan-out). Now it's a complete legacy product a family can trust.
• C2 — one-facility referral pilot. Not a license — a referral relationship: welcome-packet insert + activity-director endorsement at one community, family pays. Instrument referral → signup → capture-completed. Prove the channel.
• C3 — channel partner. An eldercare channel partner (or facility group) refers at scale, white-labeled; we run the rails. Mirrors the DreamPro/Ryan-Thomas white-label pattern.

11. Open decisions (with leans)

• Brand. Lovio, or a dedicated grief-sensitive sub-brand for eldercare? Lean: a distinct skin/brand — the facility + grief context wants its own calm, non-tech framing (the Lovio rebrand + FutureSend thinking is adjacent), reusing the Lovio engine underneath.
• Read-only vs. generative for the deceased. Lean: read-only, hard. Generative ("answer a grandchild's new question in Grandpa's voice") is a separate, deliberately-greenlit bright-line decision requiring explicit pre-death opt-in — out of scope for launch.
• Pricing unit. One-time keepsake fee vs. subscription. Lean: one-time capture package + optional cheap hosting, because the dying competitors prove families won't trust an indefinite subscription to hold the dead.
• Capture device. Family's phone (facilitated) vs. a facility kiosk. Lean: phone first (zero hardware, the child already has it).
• Counsel gate. The §7 consent/publicity architecture needs an elder-law + right-of-publicity attorney before C0 ships, not after.

12. Success metrics & kill criteria

• North star: capture-completed families (a family that finished ≥1 guided session + sealed ≥1 capsule).
• Funnel: referral/ad impression → signup → first session → capture-completed → capsule sealed → (later) delivered.
• The unproven hinge to validate first: does placement guilt actually convert to a purchase? The research found this asserted only in competitor marketing — no rigorous data. C0's real job is to prove willingness-to-pay at the placement moment. If it doesn't convert, the wedge is wrong regardless of how good the product is.
• Kill/pivot: if consent/capacity + publicity law makes the subject-consents model legally unworkable at acceptable cost (per counsel), the whole wedge is blocked — resolve this before building past C0.

13. Sources & soft spots (flagged honestly)

Grounded in a cited 2026-07-08 market brief. Load-bearing but soft claims to verify before betting on them:

• Guilt → spending conversion is asserted in marketing content only — no rigorous data. This is the #1 thing C0 must validate.
• Familiar-voice / Simulated Presence Therapy clinical evidence is graded very low — don't overclaim it; lead with reminiscence/life-review (well-evidenced) instead.
• Publicity-rights / POA-doesn't-cover-likeness / post-mortem-voice statutes come from law-firm secondary sources — get an attorney before relying (this gates C0).
• A circulating "95% depressive symptoms at 90 days" stat could not be traced to a primary source — excluded.

See also

• LIVING_VOICE_ROADMAP.md — Track C is the parent; this doc revises its business-model strawman (facility-license → adult-child wedge).
• LIGHTHOUSE.md — shares the safeguarding/consent DNA (opens-on-their-terms, practitioner-gating, never-a-push).
• AR_GLASSES_LIVING_VOICE.md — integration #7 (Legacy Channel, hands-free) is the glasses extension of this.