In the silence between visits, the pet owner is alone and the practice is blind. We put a concierge in that silence — the owner’s honest advocate, gifted by the practice, with no thumb on the scale — that meets the owner where they are, gives the practice its sight back, and becomes the shared surface where the relationship lives. It’s defensible because we originate a two-sided truth no one else has. And it could change lives because it stands with people in the moment love meets a number.
Soul + substance, current as of tonight · read alongside the story · nothing locked
01 · PART I — THE SOUL
The truth we’re building on
A practice’s relationship with a pet is a burst of attention during a visit, then months of silence. That silence is where almost everything important quietly happens — and where almost everything is lost. It has three faces, one per person standing at its edge:
For the owner — loneliness
Alone at 9pm with a scared animal and nowhere authoritative to turn, so they ask strangers on Facebook. The vet has authority but isn’t available; the group is available but has none.
For the practice — blindness
A practice doesn’t lose patients; it loses sight of them, through non-events no busy day can show you. You can’t notice who isn’t in front of you.
For both — shame
When love meets a number, the owner who can’t do “what a good owner would” goes quiet in shame — and the vet never knows they were sitting in a parking lot, frozen.
Read the story first — The Quiet Between Visits (the felt version of everything below). This document is the strategy that serves it.
02
The thesis
Bring the dignity of participation into the relationship between a pet owner and their vet. Stop designing for silence and shame; design a surface where the two do the real work of caring for an animal together — mediated by a concierge that absorbs the labor and leaves the judgment human. Everything else, including the shift from reactive to proactive care, falls out of getting this one thing right.
The reframe
Relationship, not messaging
The market sells pipes (texting + reminders). We build the shared surface those pipes were a workaround for — where the relationship lives.
The payoff that falls out
Proactive care, finally affordable
Participation turns the owner into the sensor; that’s what makes proactive care possible. It was never a will problem — it was a cost-of-attention problem. We unblock care, we don’t replace it.
Why we can
An adaptive, owned surface
The site can be different per person, per moment, and we already own it in every practice. No static site or templated tool can follow.
03
The charter — the wall the rest answers to
The concierge has no thumb on the scale. Not for the practice’s revenue. Not for the cheapest path. Its only loyalty is to the owner’s own informed, unashamed choice. It helps them see and never decides; it makes it safe to be a person with real constraints who loves their animal; it is honest about risk and unashamed about constraint at the same time; it hands them back to their vet; the heaviest moments belong to a human. When someone someday asks it to “drive conversions,” the answer is already no. (Full text: the Concierge’s Charter.)
04 · PART II — WHAT IT IS
The concierge
One presence, two faces — the steward of the relationship between owner and practice. It belongs to neither; it serves the bond. DE is invisible behind it.
To the owner — their advocate
Someone in their corner, who knows their pet, always on. Slightly separate from the practice on purpose — that’s what makes it trustworthy in the money-and-care moment. Gifted by the practice (which deepens the bond, not replaces it).
To the practice — its peripheral vision
The thing that notices who’s slipped, surfaces the lost-sight-of, and never lets anyone fall through — sight into the dark they can’t see.
Honest about what it is
Never pretends to be a person. It’s the practice’s concierge, an AI advocate, and it says so — honesty is what makes it trustworthy. A human steps in by name when it matters.
Whose it is
Brand the capability to the practice (“give every client a concierge”); white-label the instance to the owner (it’s Maple Creek’s). DE never sits between practice and owner.
05
How it works — participation, both sides
The owner self-authors their role
We read who they’re being (anxious novice, co-caregiver, “just tell me”) from behavior and meet them there — never assign a role, never ask via a form. “Let the vet decide” is a dignified role too.
The practice self-authors its role
It sets how much it shows up — voice, delegation depth, escalation, after-hours posture — with safe, conservative defaults so a solo practice gets a sane setup for free.
See → judge → execute
The AI surfaces (proactively); the human judges (the call stays theirs); the AI does the labor. Judgment, not labor — and the tender judgments are always human.
Who answers: a classification, not a toggle
Labor → the concierge. Judgment → a human. Danger / clinical / grief → a human, always. Transparent who’s acting; safety overrides everything.
06
The surface — a magic link into a real space
Portals died of friction; text won by being frictionless but stayed thin. The magic link breaks the trade: the practice (via read access) texts the owner a pre-authenticated link — no app, no login, text-level friction — that opens into a real shared space: the recovery roadmap, the photo timeline, the pet’s living record. The notification is the pulse; the magic link is the frictionless door; the space is the expansive room — far more than text could hold, and the home where the relationship accumulates.
07
The heart — the money-and-care moment
This is the moment the whole thing exists for, and the hardest thing in veterinary medicine. The concierge is a true advocate here: it helps the owner reason through their own context — the real spectrum of options, honest about what each costs them and what it could cost the pet — removes the shame, and hands them back to their vet informed and empowered to choose for themselves. No thumb on the scale. Made in shame, that decision goes badly; made clear-eyed and unashamed, it’s the right one for them — and that changes lives on both sides of the table.
08 · PART III — WHY IT’S DEFENSIBLE
The moat (or: why a horizontal AI doesn’t eat this)
The generic capability — “make your business queryable” — will be eaten by Cowork/Codex-class agents. We don’t win there. We win on what they structurally can’t get:
Origination, not aggregation
The between-visit relationship, the participation, the intake exist in no system to point an agent at. We create the data. You can’t query what hasn’t been originated.
The two-sided bridge
A horizontal agent sees one side (the practice’s data). We sit on both — owner and practice — a position a one-sided tool can’t occupy.
The fleet
Patterns only visible across hundreds of practices. A single practice pointing ChatGPT at its own data can’t see them.
The grunginess + the trust
Legacy PIMS, no APIs, VCPR, low-tech trust-driven buyers — too ugly and small for generalists, exactly the plumbing they won’t do.
So we don’t fight the horizontal agent — we become the source it must call. When an owner asks their own assistant “is this normal for Bella,” the authoritative answer is ours. The moat is: originate the two-sided truth; be the thing every agent has to come through.
09
The system — one stack, and the horizon
Three things DE is building become one continuous relationship with the pet, bracketing every visit:
live
Spotlight
Pre-visit intake → a Subjective note written back into the PIMS. The participation loop, before the visit. (Deliberately no second system.)
↓
the concierge
The shared surface
The relationship between visits — the owner’s advocate + the practice’s peripheral vision. The recovery room is the wedge.
↓
R&D
Prism + the provenance layer
Read access to PIMS truth + every fact stamped with where it came from and how sure we are — what makes the concierge trustworthy and agent-callable.
Horizon (not now): owners increasingly act through their own agent. We build behind clean endpoints + provenance so “open it to their agent” is a permissions decision, not a rebuild. The UI gets disintermediated; being the source they call only grows more valuable. Never give up ownership of the practice’s truth, data, and relationship.
10 · PART IV — THE GROUND
The practice is a real party
The binding constraint is the workforce
Not demand — people. Burnout and turnover are the #1 problem. Everything must absorb load, never add it.
No second system
They won’t adopt another login or change PIMS. Value lands in the flow they already use; the concierge eats tabs, it doesn’t add one. (Spotlight’s proven discipline.)
Route to the tech, protect the DVM
The credentialed tech is under-utilized; the DVM is the scarce resource. Surface only the ~20% that needs the doctor.
It spans the whole lifecycle
New-pet → wellness → the “is this normal” moments → surgery/recovery (the wedge) → chronic → senior → end-of-life. Different register at every stage; emergencies stay reactive-fast.
11 · PART V — THE PATH
Build sequence: substrate first, prove then scale
The safety + governance substrate, first.The clinical-guardrail layer (the AI can’t give de novo clinical advice), the governance control plane (the practice approves what it may say), closed-loop acknowledgment + practice-set SLA, and net-negative-workload + abandonment detection. “We unblock care” dies if the AI is unsafe, ungoverned, or adds work — so this is the foundation, not an appendix.
The wedge: the Recovery Room.Bounded, high-participation, email-powered today, emotionally charged. Proves the whole loop in miniature.
Extend, trunk by trunk.Chronic, senior, wellness — only after the loop is proven safe and load-negative. Never the whole lifecycle at once.
12
What we have to prove — and how we’ll know
The falsifiable core: the participation loop — owner-as-sensor → concierge → selective practice — can absorb practice load while turning owner signal into a real early-catch, safely. If that holds on one bounded wedge, the rest follows. If it doesn’t, no amount of beautiful design matters.
North-star metric
Cost-of-attention down × care-realized up. Fewer human-hours and more recommended care actually happening. If both move, we’ve solved the thing nobody could.
Refuse to measure
Session frequency / engagement-for-its-own-sake (rewards over-burdening). A quiet, healthy stretch is success.
13
From here
Consolidate the soul & get Alie all the way in.This document + the story + the charter. The people who build it have to hold it.
Touch reality — the cheapest test of the riskiest assumption.Put the story, the concierge, the charter in front of 3 real believer-vets and a handful of owners. Does it ring true to a vet? Does the advocate-in-the-money-moment move an owner, or land as creepy? Learn cheap, before building.
Then plan & build — soul-first.Substrate → wedge → prove the core → extend. The plan earns its place after reality has spoken.
How we keep faith with the vision: a thing like this is never let down by being built slowly or imperfectly — only by being quietly compromised. The charter is the wall against that. Carry the next honest step, not the whole weight.
14 · PART VI — DECISIONS
Open questions — the red-pen targets
1 ★
The money-and-care line.
How far does the concierge go in genuinely serving the owner over the practice’s short-term revenue? (Charter says: all the way. Confirm we’re committing to that — it’s the bet on good practices that becomes the moat.)
2 ★
The riskiest assumption to test first.
Is it “does it ring true to a real vet/owner” (my lean), the loop absorbing load safely, or the believer-vet existing and wanting it? Pick the one whose failure would hurt most.
3 ★
Does the concierge get a name / persona?
A named “her” makes “someone in your corner” vivid; the floor is honesty (never pass as human). Practice-branded or a DE-level persona?
4
Substrate-first, agreed?
We build the guardrail + governance before the conversational surface — slower to a demo, but the only safe order.
5
Transparency to the owner.
Confirmed: “it’s the practice’s concierge, honestly an AI, human one tap away.” Any edge we’re missing?
6
How much does Prism need to be real for v1?
The story assumes it works (auto magic-link). What’s the minimum read-access that makes the wedge true?