FetchBack — revenue-recall desk for independent vet clinics

1. One-liner

AI works the declined-estimate and overdue-recall backlog short-staffed vet clinics never get to, recapturing booked revenue.

2. Trend signal — why now?

Three things collided in the last 12 months. First, the veterinary workforce crunch got worse, not better — the US is staring at a projected shortfall of 15,000–41,000 veterinarians by 2030, and credentialed technician shortages are now the binding constraint on clinic throughput (Virginia Tech VetMed reporting, Feb 2026; NIFA shortage maps). When a two-person front desk is drowning, the first thing that gets dropped is proactive outreach: the client who declined the $1,400 dental, the dog overdue for its heartworm recheck, the cat whose senior bloodwork lapsed eight months ago. That revenue doesn’t bounce — it silently walks to the clinic down the road.

Second, the phone data quantifies how bad the leak is. Independent clinics miss 30–42% of inbound calls, caller satisfaction collapses after ~43 seconds on hold, and the overwhelming majority of callers who hit voicemail never call back — they call a competitor (AgentZap veterinary phone stats, 2026; Brilo AI small-business voice review, 2026). The same staffing math that drops inbound calls guarantees outbound follow-up never happens at all.

Third, AI agents that can actually do this work shipped cheaply this quarter. Anthropic’s May 2026 developer platform added multi-agent orchestration and self-hosted sandboxes, with API rate limits raised up to ~17×; OpenAI shipped parallel agent tooling the same week (Anthropic Dev Conference coverage, May 2026; The New Stack, May 2026). An agent that reads a practice-management export, drafts a per-pet, per-owner message in the vet’s voice, and books the slot is now a weekend of plumbing, not a research project.

Provenance:

• Signal 1 (demand): Independent vet clinics miss 30–42% of calls, callers don’t retry; staffing shortage means outbound follow-up never happens — AgentZap veterinary phone statistics / Brilo AI — https://agentzap.ai/blog/veterinary-phone-statistics — 2026-05
• Signal 2 (economic/structural): US vet workforce shortfall projected 15,000–41,000 by 2030; tech shortage is the throughput bottleneck — Virginia Tech VetMed / NIFA shortage maps — https://news.vt.edu/articles/2026/02/vetmed-vth-veterinary-workforce.html — 2026-02
• Signal 3 (feasibility): Cheap, capable AI agent orchestration shipped May 2026 (Anthropic multi-agent platform, OpenAI agent tooling, rate limits up to ~17×) — Anthropic 2026 Dev Conference / The New Stack — https://thenewstack.io/anthropic-openai-wall-street-ai-agents-developers/ — 2026-05 Category: Underserved niche

3. The opportunity

Every existing vet-comms tool — PetDesk, Vetstoria, Televet, Vetsource ScriptRight — is built around the inbound and the scheduled: appointment reminders, online booking, refill requests, two-way texting. They fire a generic “Fluffy is due for a visit” blast and call it engagement. None of them work the two highest-value backlogs:

1. Declined estimates. The owner who said “let me think about it” on a $1,400 dental or a $900 mass removal. That’s pre-qualified, already-diagnosed revenue sitting in the chart. Nobody follows up because following up well means re-reading the record, understanding why it was declined, and writing something that isn’t a coupon spam.
2. Lapsed/overdue patients past the reminder window. Generic reminder tools give up after the canned sequence. The pet that’s now 8 months overdue is invisible.

A focused AI does this 10× better than a blast tool because the work is judgment, not scheduling. It reads the clinical note, sees “owner cost-concerned, recommended dental grade 3,” and drafts: “Dr. Patel wanted to follow up on Bella’s dental — her grade 3 tartar can progress to tooth loss. We can split this across two visits or set up CareCredit; here are three openings next week.” That is the message a great front-desk lead would write if they had four hours nobody has.

4. Target market

• Primary customer: Owner or practice manager of an independent / small-group US companion-animal clinic — 1 to 4 doctors, not corporate-owned (not Mars/VCA/NVA). Roughly $0.8M–$4M annual revenue.
• Why they buy: “We know there’s money in those declined estimates and overdue patients, we just never get to it. My techs are maxed; nobody has time to chase.” It’s not a cost they feel — it’s revenue they can see leaking and can’t plug.
• Rough TAM reasoning: ~28,000–30,000 companion-animal veterinary establishments in the US; majority still independent or small-group despite consolidation. Even 3,000 paying clinics is a serious business.
• Why now for them: Staffing is worse than ever, corporate consolidators are squeezing independents on price, and recapturing even 10 declined estimates a month at ~$600 average is $6,000/month found money — the exact margin an independent needs to stay independent.

5. Product sketch (MVP)

• One-time read of the clinic’s practice-management export (declined line items, overdue reminders, last-visit dates, clinical notes).
• Auto-built worklist ranked by dollars recoverable × likelihood, not alphabetically.
• AI drafts a per-pet, per-owner outreach message that references the actual recommendation and the likely objection (cost, time, “wait and see”).
• Vet/manager approves or one-tap edits before anything sends — nothing goes out unreviewed in v1.
• Sends via SMS + email; replies route to a simple inbox with suggested responses and a booking link.
• Monthly “recaptured revenue” report: estimates re-accepted, lapsed pets rebooked, dollars attributed.
• Optional outbound AI voice call for the high-value declines that don’t answer text.

6. AI angle — what’s load-bearing

Remove the AI and you’re left with a spreadsheet and a person who doesn’t exist. The load-bearing work is (a) reading messy free-text clinical notes to extract what was recommended and why it was declined, (b) ranking the backlog by recoverable dollars, and (c) composing outreach that sounds like the clinic and addresses the specific objection. Generic mail-merge (“you’re due!”) is exactly what the incumbents already do and exactly what gets ignored. The judgment-per-message is the product.

7. Localization angle (if any)

N/A — this is a US-first play. The wedge is US practice-management data formats (Cornerstone, AVImark, ezyVet exports), US payment norms (CareCredit, Scratchpay), and US clinic economics. A localized version could follow in the UK/AU/Canada later, but geography isn’t the wedge — the unserved workflow is.

8. Business model — path to $1M–$5M ARR

• Pricing: $299–$699/mo per clinic by clinic size, with an optional performance kicker (e.g., flat tier + small % of attributed recaptured revenue for clinics that prefer it). Land on flat-fee for simplicity; the ROI story sells itself.
• ACV: ~$5,400/clinic/year at the $449 mid-tier.
• Rough math to $1M ARR: ~185 clinics × $449/mo × 12 ≈ $1.0M. Very reachable in a 28K-clinic market.
• Rough math to $5M ARR: ~930 clinics, or fewer clinics + the voice-call upsell + multi-location group deals (a 4-doctor group counts as one logo but pays more).
• Expansion path: add outbound voice for high-value declines; add refill-recapture and bonded-client win-back; sell a multi-location “group console” to small consolidators; eventually expand to dental and physio practices with the same declined-estimate dynamic.

9. Go-to-market wedge — first 100 customers

• Free “recoverable revenue audit.” Clinic sends a de-identified PMS export; we return a one-page number: “You have $38,400 in declined estimates and 210 overdue patients from the last 12 months.” That number is the sales pitch. This is the entire wedge — show them the leak before asking for a dollar.
• Practice-manager Facebook groups + VHMA (Veterinary Hospital Managers Association). Practice managers congregate in a handful of large private FB groups and at VHMA — drop the free-audit offer there, where the buyer literally manages this backlog.
• Cold outreach to independents via state VMA directories. Scrape state veterinary association member lists (independents, exclude corporate-owned), send the free-audit offer with a personalized recoverable-dollars estimate. Expect single-digit % reply on a number that specific.
• Reps/consultants who advise independents. Veterinary practice consultants and CareCredit/Scratchpay reps already sit in front of these owners talking about revenue — co-sell.

10. Build complexity — justification

Medium. The AI drafting and ranking is off-the-shelf model work. The real work is PMS data ingestion — Cornerstone, AVImark, ezyVet, Pulse all export differently, and many independents run on-prem legacy systems where the cleanest path is a guided CSV/export upload rather than a live API. A small team ships a usable v1 (upload → worklist → approve → send → report) in ~10–14 weeks; live PMS integrations come later. No custom models, no regulated PHI equivalent (animal records aren’t HIPAA-covered), no hardware.

11. Gating checklist

Gate	Pass?	Note
Legal in target market	✅	Animal health records aren’t HIPAA-covered; TCPA applies to outbound — handle consent/opt-out properly.
Ethical — no harm / dark patterns	✅	Vet approves every message in v1; no fear-mongering, opt-out honored.
Market exists (evidence above)	✅	Missed-revenue + staffing data triangulates; incumbents prove willingness to pay for comms.
1–5 person team can build this	✅	Medium build, off-the-shelf AI, no regulated infra.
Launchable with <$50K / ₹40L	✅	Solo/pair + inference costs; free-audit GTM is near-zero CAC.

12. Feasibility score

Axis	Weight	Score	Notes
Problem intensity	20	16/20	Visible, recurring revenue leak; staffing crisis guarantees it stays unworked. Not quite hair-on-fire (it’s lost upside, not an active bleed).
Demand evidence	15	12/15	Strong adjacent signals (missed-call data, incumbent willingness-to-pay) but I lack a direct “I’d pay for declined-estimate recapture specifically” quote — verify in sprint.
Build feasibility	15	11/15	AI is easy; messy PMS export ingestion across legacy systems is the real work.
Distribution clarity	15	11/15	Free-audit wedge is concrete and the buyer is clustered (VHMA, FB groups, state VMAs); conversion math still unproven.
Revenue mechanics	15	12/15	Clear ROI story, benchmarked pricing vs PetDesk-class tools; ACV solid.
Time to first revenue	10	7/10	Free audit → paid in weeks once the audit number lands; needs the audit pipeline built first.
Defensibility	10	5/10	Execution + niche workflow lock-in (the recapture data compounds), but copyable by an incumbent that decides to build it.
Total	100	74/100

13. Qualitative modifiers

Founder-fit tags

technical-heavy · domain-expertise-required

Key assumptions to validate (3–5)

1. Assumption: Independents have a meaningful, quantifiable backlog of declined estimates + overdue patients in their PMS. How to test: Run 10 free audits on real exports; measure average recoverable dollars. Kill if median < $15K/clinic/yr.
2. Assumption: Owners will pay $300–700/mo to recover it. How to test: Take the 10 audit numbers to those owners; count how many sign a paid pilot at full price.
3. Assumption: AI-drafted, vet-approved outreach actually converts declines/lapsed pets (not just sends). How to test: Run one clinic’s backlog for 60 days; measure re-accepted estimates and rebookings vs. their baseline.
4. Assumption: PMS export ingestion is tractable across the top 3–4 systems without per-clinic custom work. How to test: Pull real exports from Cornerstone, AVImark, ezyVet; confirm one parser handles each.

Risk flags

1. Platform/competitive risk: PetDesk, Vetsource, or a PMS vendor (ezyVet/Covetrus) bolts declined-estimate recapture onto an existing install base and undercuts on bundling. This is the top threat.
2. TCPA / consent risk: Outbound SMS to clients needs clean consent and opt-out plumbing; sloppy execution invites complaints. Lean on existing client-consent and vet approval.
3. Data-access friction: Legacy on-prem PMS with locked exports could make onboarding painful enough to throttle growth — mitigate with dead-simple guided upload.
4. Attribution disputes: Clinics may argue a rebooking “would have happened anyway,” pressuring any performance-fee model. Default to flat fee to avoid the fight.

14. Structured verdict

Score:                  74/100
Verdict:                GO
Confidence:             Medium
Best-fit builder:       Technical founder + a veterinary practice-manager advisor
Time to revenue:        6–10 weeks (free audit → paid pilot)
Capital to launch:      $8–15K ($ inference + data plumbing + a part-time vet advisor)
Top 3 assumptions to validate first:
  1. Median recoverable backlog ≥ $15K/clinic/yr — run 10 free audits on real PMS exports
  2. Owners pay $300–700/mo once shown their number — convert audits to paid pilots
  3. AI-drafted vet-approved outreach actually re-books — 60-day backlog run at one clinic vs baseline
Kill criteria:
  - Abandon if median audit recoverable backlog < $15K/clinic/yr across 10 clinics
  - Abandon if <20% of clinics shown a 5-figure audit number sign a paid pilot
  - Abandon if a PMS vendor or PetDesk ships native declined-estimate recapture before v1

15. Next step — 1-week validation sprint

• Day 1–2: Recruit 8–10 independent clinics from one state VMA directory + a practice-manager FB group with a single offer: “Send me a de-identified export, I’ll tell you how much recoverable revenue is sitting in your declined estimates and overdue patients — free.”
• Day 3–4: Build the throwaway audit script, run the exports, produce a one-page recoverable-dollars number per clinic. Note how clean/dirty each PMS export is.
• Day 5: Take each number back to the owner and ask for a $449/mo 60-day paid pilot. Go/no-go: ≥3 of 10 say yes at full price and median backlog ≥ $15K/clinic/yr. Anything less, the leak isn’t big enough or owners won’t pay to plug it — kill or re-scope to a different vet revenue leak.

The result is falsifiable: a paid-pilot count and a median dollar figure, not a vibe.