Case file — 92EBD646
The idea
“Physical therapists spend 30-40% of their time on SOAP note documentation required for Medicare/Medicaid billing. AI medical scribe tools (Nuance DAX, Abridge, Suki) are built for physicians and don't understand PT-specific clinical vocabulary, outcome measures (LEFS, NPRS, DASH scores), or the functional limitation language required by CMS for PT reimbursement. Dragon Medical One ($1,200-2,400/year) is expensive and generic. WebPT and Clinicient (the dominant PT EHRs) have basic documentation templates but no AI generation. A mobile app that listens to PT treatment sessions, auto-generates compliant SOAP notes in CMS format with PT-specific vocabulary and automatically populates outcome measure scores, then exports directly to WebPT or Clinicient. Price: $100-200/month per therapist. 250,000+ PTs in the US, most in private practice with full budget autonomy.”
The panel
# Market Analysis: PT-Specific AI Scribe No live data exists for PT-specific AI scribes or direct competitors named in your search results—only generic mentions of DeepCura ($129/mo, PT templates) and ScribePT (AI documentation for PT). The live search confirms the core problem: PTs document 8–15 minutes per encounter, and general physician tools miss PT-specific vocabulary and CMS compliance requirements. Market is clearly growing (multiple tools emerging), not saturated. Red flag: CMS reimbursement rules for PT documentation are hyperlocal by state, payer, and diagnosis code. A national app risks building compliance that fails in key markets, requiring constant legal/billing audits. Founders often underestimate this friction. Genuine strength: 180,000+ private-practice PTs with direct budget autonomy and no dominant AI solution yet—you're pre-category-creation. If you nail WebPT/Clinicient API integration and CMS compliance in one state, word-of-mouth scales fast in a tight professional network.
Core underestimation: You're assuming audio-to-compliant-documentation is a solved problem. It isn't. PT sessions involve spatial positioning, movement quality, and functional demonstration that audio alone cannot capture. A therapist saying "patient demonstrated improved dorsiflexion" means nothing without knowing the baseline, resistance applied, or compensation patterns. Your model will generate plausible-sounding notes that fail audit because they lack the clinical specificity CMS demands. You'll need video + pose detection or therapist manual verification, which destroys your unit economics. Build-vs-buy trap: Don't build EHR connectors to WebPT/Clinicient yourself. Their APIs are unstable, they actively resist third-party integrations, and a single breaking change kills your product. You'll spend 40% of engineering on integration maintenance. Pivot to a standalone note export (PDF/HL7) that therapists copy-paste—unsexy but survivable. Moat: Weak. Once Nuance/Abridge recognize PT as a segment, they'll retrain models in 6 months. Your only defensible moat is outcome-measure extraction, but that's a feature, not a product. What works: PT-specific vocabulary training data is genuinely scarce and valuable. If you can license de-identified notes from WebPT or clinics, you own a real dataset advantage for 18–24 months.
Your CAC/LTV problem is distribution. PTs don't buy software through app stores or online ads—they adopt via EHR integrations, word-of-mouth within clinics, or rep relationships. You're pricing at $100-200/month ($1,200-2,400 annually), but your actual CAC will likely be $800-1,500 per customer if you're cold-calling practices or relying on digital channels. LTV math only works if churn stays under 5% monthly; one bad export to WebPT or Clinicient kills retention instantly. Your pricing assumption that's probably wrong: you're anchoring to Dragon Medical's $1,200-2,400 annual cost, but PTs see that as a sunk expense they already own or have rationalized away. You're asking them to add a new subscription on top of their EHR bill. Adoption accelerates only if you embed directly into WebPT/Clinicient, which requires their partnership—and they may build this feature themselves within 18 months. At zero traction, you have maybe 18-24 months of runway before you need paying customers if bootstrapped, or 12 months if VC-funded. The real constraint is product-market fit validation: you need 20-30 paying practices generating clean, compliant notes and actively renewing before you know the model scales. One thing working in your favor: CMS compliance is a genuine pain point with real financial consequences (denied claims, audit risk). If your notes reduce claim denials by even 3-5%, the ROI conversation becomes easy. That's defensible pricing power.
Timing verdict: Late, but with a narrow opening. AI medical scribe adoption in PT is ~18 months behind physician adoption—the market has moved from "does this exist?" to "why isn't my EHR doing this?" WebPT and Clinicient know this gap exists and have the distribution to fill it themselves. Your window closes when they bundle native AI note generation into their platforms, which is likely 12–18 months away given their parent company resources and the simplicity of PT-specific training data. Macro trend: CMS documentation audits on PT claims are intensifying post-2024 telehealth expansion. Compliance pressure is real, but it's also pushing the big EHRs to move faster on this exact feature. Opportunity window: Closing. You're not too early—the pain is proven—but you're arriving as incumbents recognize the same problem. One genuine timing advantage: Private practice PTs are desperate now and will pay directly for relief before waiting for EHR updates. That 12–18 month window is real money if you can get 500–1,000 early adopters before the incumbents ship.
Competitors found during analysis
Live dataDeepCura
not stated raised
$129/mo, PT templates included
ScribePT
not stated raised
AI docs, PT-exclusive focus
Cause of death
Audio-only capture can't survive a CMS audit
Your CTO-equivalent panel member flagged this and it's the most dangerous assumption in the pitch. PT sessions involve spatial positioning, movement quality, resistance levels, and compensation patterns that audio alone cannot capture. A therapist saying "improved dorsiflexion" generates a plausible-sounding note that lacks the clinical specificity CMS auditors demand — baseline measurements, degrees of motion, resistance applied, functional context. You'll either need video/pose detection (which explodes your build complexity and privacy requirements) or heavy manual therapist verification after every session (which destroys the core value proposition of saving 30-40% of their time). You're promising automation but delivering a fancy dictation tool with a verification step.
The EHR integration trap will eat your engineering budget alive
WebPT and Clinicient APIs are unstable, and both companies have strategic incentive to resist third-party integrations that could be replaced by their own native features. Your tech panel estimates 40% of engineering time on integration maintenance alone. But here's the deeper problem: without seamless EHR integration, your product is a copy-paste workflow — and PTs already have a copy-paste workflow. It's called typing. The integration IS the product for most buyers, and it's the one thing you can't reliably control.
You're racing a closing window with zero velocity
The timing panel puts your window at 12-18 months before WebPT or Clinicient bundles native AI note generation. You're at the idea stage. Assume 3-4 months to build an MVP, 2-3 months to get your first 10 paying clinics, another 3-4 months to iterate on compliance accuracy. That's 8-11 months before you even know if the product works — leaving you maybe 3-6 months of clear air before incumbents ship. And they don't need to ship something great. They just need to ship something "good enough and free" bundled into the EHR subscription PTs already pay for.
⚠ Blind spot
You're framing this as a software problem, but it's actually a compliance liability problem. The moment your AI-generated note causes a denied claim or triggers a CMS audit, you don't just lose that customer — you lose their entire referral network. PTs talk. In a professional community of 180,000 where word-of-mouth is the primary distribution channel, one compliance failure at the wrong practice cascades into a reputation problem that no amount of marketing fixes. And CMS compliance rules vary by state, payer, and diagnosis code — meaning your "national app" is actually 50 different compliance products duct-taped together. You need a compliance team before you need a sales team, and that's not in your cost model.
What would need to be true
PT-generated audio must contain enough clinical specificity — without video or manual input — to produce notes that survive CMS audit at a 95%+ accuracy rate, which no current audio-only medical scribe has demonstrated for any movement-based specialty.
WebPT and Clinicient must fail to ship native AI documentation features for at least 18 months, or must ship versions bad enough that PTs actively seek alternatives — a bet against well-resourced incumbents solving a problem they already know exists.
Your compliance accuracy must be near-perfect from day one in at least 3-5 high-volume states, because a single wave of denied claims in a tight-knit professional community will poison your distribution channel faster than any marketing can repair it.
Recommended intervention
Stop building a mobile scribe. Build a CMS compliance verification engine instead. Position the product not as "we write your notes" but as "we audit your notes before CMS does." Take the notes PTs are already writing — whether by hand, dictation, or template — and run them through an AI layer that flags missing functional limitation language, incomplete outcome measures, and payer-specific documentation gaps. This flips every problem in your favor: you don't need audio capture (eliminating the spatial/movement problem), you don't need deep EHR integration (you can ingest exported notes or screenshots), you reduce compliance liability (you're flagging errors, not generating them), and you become more valuable when WebPT ships native AI notes (because someone still needs to verify those notes pass audit). Price it at $79/month as insurance against denied claims. The ROI pitch writes itself: "How many claims did you have denied last quarter? We cost less than one."
Intervention unlocking
5seconds
No account needed. One email, no follow-ups.
Want your idea examined? Free triage or full panel →