ChargeSweep — missed-charge recovery for legacy-PIMS vet clinics

1. One-liner

Reads your day’s medical records against the invoices and catches the exams, labs and meds nobody billed.

2. Trend signal — why now?

Three things landed in the same 12 months.

The pain got measured and got worse. The AVMA pegs missed charges at 5–10% of revenue. The 2025 VHMA audit survey found inaccurate billing was a top-two finding in 51% of practices and incomplete documentation in 54%. A multi-doctor hospital audit found 5–20% of charges unbilled — roughly $500K/year. And for the first time in several years, bottom-line profits fell at US veterinary practices in 2025. The leak was always there; the margin to absorb it is gone.

The tech to read a vet record cheaply arrived. AI scribe and ambient charge-capture went from demo to production in 2025. Instinct reports hospitals seeing “at least 6% increase in captured revenue” after switching to its integrated charge capture. Lupa raised $25M (Series A Oct 2025, reportedly 50x revenue growth between rounds) building an AI-native PIMS. ScribbleVet was acquired by Instinct Science in Jan 2026. “Veterinary AI” search volume grew ~1,680% YoY 2024→2025.

But the money chasing this is all rip-and-replace. Every funded player — Lupa, Instinct, Digitail, Shepherd — wants you to replace your PIMS. ~9,000 clinics run AVImark and Cornerstone is ~20% of the rest. Server-based legacy. These owners will not migrate twenty years of records to capture 7% — but they’d happily pay to plug the leak without touching their PIMS.

Provenance:

• Signal 1 (demand): AVMA 5–10% missed-charge revenue loss; 2025 VHMA survey 51% inaccurate billing, 54% incomplete docs; US vet profits fell 2025 — Today’s Veterinary Business, https://todaysveterinarybusiness.com/missed-charges-0226/ — 2026-02
• Signal 2 (feasibility): Ambient charge capture in production — Instinct reports ≥6% captured-revenue lift post-switch; ScribbleVet acquired by Instinct Science Jan 2026 — https://www.shepherd.vet/blog/8-best-ai-powered-veterinary-practice-management-software-platforms-2026-comparison-guide/ — 2026-01
• Signal 3 (economic): Lupa $25M raised, Series A Oct 2025, ~50x revenue growth; “veterinary AI” search +1,680% YoY — https://www.co.vet/post/veterinary-ai-scribe — 2026-01 Category: Tech-unlock

3. The opportunity

The whole funded field made the same bet: own the entire clinic OS. That bet works on greenfield and cloud-curious practices. It strands the largest, stickiest segment — independent practices on AVImark/Cornerstone — because the switching cost (data migration, retraining a 12-person team, downtime) dwarfs the 7% they’d recover.

The 10× move is to not be the PIMS. ChargeSweep is a read-only overlay: it ingests the day’s finalized medical records and the day’s invoices from whatever PIMS the clinic already runs, and within minutes flags every documented service that has no matching line item — “Bella’s chart says senior wellness panel drawn, fecal run, gabapentin dispensed; invoice has exam + vaccine only.” A practice manager clears the worklist before the client leaves or, worst case, before the day closes.

Incumbents do this badly because their charge capture only works inside their own ecosystem. Vetsource, dvm360 and every PIMS blog tell practices to “spot-check 10 charts per provider per week” manually — an admission that comprehensive audit is too labor-intensive to do by hand. That manual gap is the wedge.

4. Target market

• Primary customer: Owner or practice manager of an independent US small-animal practice, 1–6 vets, $700K–$4M annual revenue, running AVImark, Cornerstone, or Impromed (server-based), no in-house RCM staff.
• Why they buy: In their words — “We did much better yesterday. We only missed a six-month heartworm preventive, one fecal test, and a senior blood panel.” (practice owner, Today’s Veterinary Business). That’s ~$400 in one day they only caught because someone happened to look. They know it’s happening; they have no systematic, low-effort way to stop it short of buying a whole new PIMS.
• Rough TAM reasoning: ~28,000–30,000 US companion-animal practices; AVImark alone ~9,000 customers, Cornerstone large share of the rest. Conservatively 12,000–15,000 independent practices on legacy server PIMS. At $4,800/yr ACV, even 5% penetration ≈ $3M+ ARR.
• Why now for them: 2025 was the first profit-down year in memory. The owner who shrugged at “7% leakage” while practices grew 10%/yr is now staring at a flat P&L and a recovery that pays for itself in week one.

5. Product sketch (MVP)

• Daily import of finalized medical records + invoices from the clinic’s existing PIMS (export file, scheduled DB read, or PIMS API where one exists)
• AI cross-check: every documented service, diagnostic, and dispensed med matched against invoice line items
• Discrepancy worklist: ranked by dollar value, one click to mark “billed / waived / not applicable,” with the chart snippet shown inline
• Same-day “catch-it-at-checkout” mode: morning’s records swept by lunch so front desk can recover before clients leave
• Weekly leakage report by provider and category (the number that makes the owner renew)
• Recovery dashboard: dollars flagged vs. dollars recovered, cumulative — the ROI receipt
• No writes to the PIMS. Read-only by design (removes the #1 IT objection)

6. AI angle — what’s load-bearing

The entire product is an LLM reading free-text and semi-structured veterinary records and reconciling clinical intent against a billing ledger. “Doctor noted ‘started on Apoquel 16mg, dispensed 30 tabs’” → does an Apoquel line exist on the invoice at the right quantity? That’s entity extraction + clinical-synonym normalization (a SOAP note says “ED&C,” the price list says “mass removal”) + reconciliation logic, per record, at clinic scale. Remove the AI and you’re back to a human spot-checking 10 charts a week — which is exactly the status quo this replaces. AI is the product, not a feature.

7. Localization angle (if any)

N/A — this is a US-first play. The wedge is structural to the US market: a long legacy-PIMS tail (AVImark/Cornerstone) plus high per-visit revenue that makes 7% leakage worth four figures a month. The same idea is weaker in markets with lower ticket sizes and more cloud-native PIMS adoption. US-only is the right call, not a missing angle.

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

• Pricing: $399/mo single-doctor, $599/mo 2–4 doctor, $899/mo 5+ doctor. Flat per-practice, not per-seat (per-seat punishes the exact behavior — more staff touching records — that causes the leak).
• ACV: ~$5,800 blended.
• Rough math to $1M ARR: ~170 practices × ~$5,800 = ~$1M. ~170 of 12,000+ target practices = 1.4% penetration.
• Rough math to $5M ARR: ~860 practices (~6% of the legacy-PIMS segment), plus an upsell tier (controlled-drug log reconciliation, inventory shrink detection — same record-reading engine, new worklist).
• Expansion path: Land on missed-charge recovery; expand into adjacent record-vs-reality audits (Rx vs. dispensed controlled-substance logs, lab-machine run logs vs. invoices, inventory depletion vs. billed). Each is a new module on the same ingestion pipe; ACV climbs to $10K+ without adding seats.

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

• ROI-on-a-plate cold outreach: Run a free 5-day retroactive sweep. Practice exports last month’s records + invoices; we hand back “you left $9,400 on the table in 30 days, here are the 140 charts.” That report is the sales pitch. Target list: scrape the AVMA / state veterinary board directories + AVImark/Cornerstone user forums and Facebook groups (e.g., “Veterinary Practice Managers” ~30K+ members).
• Practice-manager communities: VHMA (Veterinary Hospital Managers Association) members are the buyer and the influencer. Sponsor/present a “what your missed-charge number actually is” session; offer free sweeps to attendees.
• Consolidator & buy-side angle: Vet practice acquirers run charge-capture audits during diligence anyway. Sell ChargeSweep as the post-acquisition tool to lift recovered revenue across a portfolio of newly-bought independents still on legacy PIMS.
• CPA / practice-consultant channel: Veterinary-specialist CPA firms (KSM, VMG-aligned consultants) already advise clients on missed charges; referral fee for every practice they route.

The pitch closes itself because the free sweep produces a falsifiable dollar number tied to the prospect’s own records.

10. Build complexity — justification

Medium. The AI reconciliation on off-the-shelf models is tractable and most of the value. The real work is ingestion from legacy server PIMS: AVImark/Cornerstone have no clean public API, so v1 leans on scheduled record/invoice exports (and read-only DB connections where IT allows). That’s integration grind, not research — a small team with one engineer who’ll go deep on AVImark’s data layout ships a credible v1 in ~3–4 months. Clinical-synonym normalization needs a vet advisor but not a custom model.

11. Gating checklist

Gate	Pass?	Note
Legal in target market	✅	Read-only audit of the clinic’s own records; clinic is data controller. No PHI/HIPAA (animal records).
Ethical — no harm / dark patterns	✅	Recovers revenue for legitimately performed work; flags, doesn’t auto-bill. Improves documentation.
Market exists (evidence above)	✅	AVMA/VHMA quantified loss, owner quotes, funded incumbents in adjacent space.
1–5 person team can build this	✅	1 AI/integration engineer + 1 vet-domain advisor + founder for GTM.
Launchable with <$50K / ₹40L	✅	Inference + infra + a few legacy-PIMS test environments. Well under $50K.

All five pass.

12. Feasibility score

Axis	Weight	Score	Notes
Problem intensity	20	17/20	Quantified daily money loss, owner already feels it, 2025 profit squeeze makes it hair-on-fire. Not 20: it’s tolerated until it isn’t.
Demand evidence	15	13/15	AVMA %, VHMA survey, owner verbatims, ≥6% lift claims from incumbents, heavy funding nearby. Direct demand for a standalone tool is inferred, not yet proven.
Build feasibility	15	11/15	AI part easy; legacy-PIMS ingestion is real integration grind with no clean API. 3–4 months, not 4–6 weeks.
Distribution clarity	15	12/15	Free retroactive sweep = falsifiable ROI pitch; VHMA + FB groups + consolidators are named, reachable channels. Cold conversion math still unproven.
Revenue mechanics	15	13/15	Pricing trivially justified by recovered $; flat per-practice clean; $1M needs only ~170 clinics.
Time to first revenue	10	7/10	Free sweep can convert in weeks, but ingestion plumbing per PIMS delays a smooth funnel. 6–10 weeks to first paid.
Defensibility	10	6/10	Soft moat: accumulated PIMS-export quirk handling + clinical-synonym dictionary + workflow lock-in. Funded incumbent could bolt this on, but they’re committed to rip-and-replace.
Total	100	76/100

13. Qualitative modifiers

Founder-fit tags

technical-heavy · domain-expertise-required

Key assumptions to validate (3–5)

1. Assumption: Independent legacy-PIMS practices will pay $400–900/mo for recovery without changing their PIMS. How to test: Free retroactive sweeps for 15 practices; measure how many convert to paid at quoted price within 30 days.
2. Assumption: AVImark/Cornerstone records can be reliably exported or read often enough for same-day catch. How to test: Build ingestion against 3 real practice environments (one each AVImark, Cornerstone, Impromed); confirm daily cadence is feasible without IT heroics.
3. Assumption: AI reconciliation hits a precision/recall good enough that the worklist is trusted, not ignored. How to test: Run against hand-audited records from 5 practices; target ≥90% precision on flagged misses with manageable false-positive rate.
4. Assumption: Recovered dollars materially exceed price (clear ROI). How to test: In the free sweeps, measure flagged-then-recovered $ vs. proposed subscription; need ≥5× in month one.

Risk flags

1. Platform dependency: Ingestion depends on legacy PIMS export/DB access; a vendor could restrict it. Mitigation: read-only exports are owner-controlled data; multiple ingestion paths.
2. Incumbent encroachment: A funded player (Instinct/Lupa) could ship a non-rip-and-replace overlay. Their org incentive is to sell the full platform — but watch for it; it’s the main kill risk.
3. Trust/false-positive risk: A noisy worklist gets switched off in week two. Precision is existential, not a nice-to-have.
4. Market timing: If legacy-PIMS clinics migrate to cloud PIMS faster than expected, the strandeded-segment thesis shrinks. (Migration is historically very slow — low probability, high impact.)

14. Structured verdict

Score:                  76/100
Verdict:                GO
Confidence:             Medium
Best-fit builder:       Technical founder who'll go deep on AVImark/Cornerstone data layouts, with a practicing-vet or VHMA-credentialed advisor
Time to revenue:        6–10 weeks (free sweep → paid)
Capital to launch:      $15–30K ($ inference + infra + legacy-PIMS test setups + advisor)
Top 3 assumptions to validate first:
  1. Legacy-PIMS practices pay $400–900/mo for overlay recovery — 15 free sweeps, measure 30-day conversion
  2. Daily record/invoice ingestion from AVImark/Cornerstone is feasible without IT heroics — build against 3 real environments
  3. Reconciliation precision ≥90% so the worklist is trusted — validate against hand-audited records from 5 practices
Kill criteria:
  - Abandon if <3 of 15 free-sweep practices convert to paid within 30 days
  - Abandon if reliable daily ingestion proves impossible for both AVImark and Cornerstone without per-practice custom IT work
  - Abandon if a funded incumbent ships a credible read-only, no-migration overlay before your v1

15. Next step — 1-week validation sprint

• Day 1–2: Recruit 8–10 independent practices on AVImark/Cornerstone (VHMA forums, vet practice-manager FB groups). Get one to share a real month of de-identified records + invoices.
• Day 3–4: Hand-build the reconciliation on that real data. Produce the leakage report: dollars missed, charts, by provider. Show it to 5 owners/managers.
• Day 5: Decide go/no-go on a falsifiable bar: ≥3 of 5 say “I would pay $500+/mo for this today” AND the sample’s flagged-recoverable revenue is ≥5× the monthly price. Below either bar → no-go or rework segment/pricing.

The output is a dollar number from the prospect’s own records and a written price commitment — not a vibe.