AI employees for physical therapy clinics
Stop the auth, billing, and no-show leaks draining revenue at every PT location.
Visit caps get hit before anyone notices. Denials sit unworked past the filing window. No-show slots go unfilled while the waitlist waits. Add more locations and the hole gets wider every month.
Relay builds AI employees that work on top of the EHR your clinic already runs: WebPT, Raintree, Prompt, or another. The AI handles high-volume back-office work: prior auth, billing, denials, and scheduling across every location. Your staff finalizes every action.
Sensory Speech & Occupational Therapy, a multi-location pediatric therapy group, ran two AI employees on top of their existing EHR and saw 100% claim accuracy, staff 33% more productive, claim denials down 12%, and faster documentation turnaround. In a free 30-minute intro call, we show you exactly where your practice is leaking.
Where your physical therapy clinic is leaking revenue.
Seven places the back office leaks at every location. Open the ones that sound like your clinic.
A Relay AI employee monitors every active authorization against the live schedule. When a patient is a few appointments from the cap, it drafts the re-authorization packet, diagnosis, functional goals, and clinical justification pulled from the EHR, and queues it for staff review before the cap is exceeded. It tracks submission status by payer and follows up on pending requests. The AI never submits to the payer without your sign-off, which means expired authorizations stop turning into write-offs.
Most commercial payers require prior authorization for a PT episode, and approved visit counts vary widely by plan. UnitedHealthcare, Cigna, Aetna, and BCBS plans all have their own rules. WebPT, Raintree, and Prompt track the visit count, but none alert you before it runs out. When the therapist delivers the visit, the authorization is already spent and the claim comes back denied.
The KX modifier signals an exception to the annual Medicare therapy threshold; visits past the threshold without it are denied. The GP modifier identifies services under an outpatient physical therapy plan of care. The AI employee validates both on every outgoing claim. It reads the auth data WebPT, Raintree, or Prompt already holds and closes the gap those systems leave to your front desk.
A Relay AI employee runs a pre-submission check on each outgoing claim. It checks required modifiers by payer and CPT code, verifies time-based codes against the 8-minute rule, and cross-references the authorization on file. Claims that need a fix are flagged for a billing coordinator to correct before submission, while clean claims pass straight through. That way, coding errors get caught before they become denials.
Most PT EHRs do basic claim scrubbing for missing NPIs and invalid diagnosis codes, but payer-specific rules are not pre-tuned. The right modifier for a supervised modality, the KX modifier for the Medicare therapy threshold, the 8-minute rule for time-based codes: a claim can pass the EHR scrubber and still fail at the payer. WebPT cites 67% of denials as stemming from front-office or documentation errors, many of them modifier or coding issues a pre-submission check would catch.
The 8-minute rule governs how time-based therapy CPT codes (for example 97110, 97140) convert treatment minutes into billable units under Medicare. The AI employee applies it automatically before the claim leaves your system. This runs on top of your existing billing workflow with no new platform to manage.
A Relay AI employee reads each denial remittance and groups denials by root cause: auth missing, modifier incorrect, timely filing, eligibility lapse, medical necessity. It drafts the corrected claim or appeal letter with the relevant note and auth attached, and routes it to the billing coordinator for one-click resubmission. It works the aging queue daily so nothing ages past the filing window unworked.
Denials come back as CARC and RARC codes that are hard to decode without a specialist. The appeal work (pulling the auth, pulling the note, writing the letter, refiling) gets deprioritized whenever the team is busy with live patients. Every denial that ages past the timely-filing window is revenue written off permanently.
Underpayments are part of the same back-office leak. The AI reconciles posted ERAs against contracted rates by payer and CPT code, flags short-paid claims for a billing lead to dispute, and surfaces patient balances that cleared insurance but were never billed out. That review runs daily, not quarterly. WebPT attributes 67% of denials to front-office or documentation errors. The AI employees described here target the modifier and coding errors that drive that 67% before the claim ever leaves your system.
Every denial that ages past the timely-filing window is revenue written off permanently.
A Relay AI employee runs batch eligibility sweeps against the upcoming schedule each evening. It flags coverage lapses, deductible resets, and secondary coordination-of-benefits changes to the front desk before the visit. Staff review flagged patients and decide how to proceed. It works on top of Availity and the EHR you already run. Stale eligibility gets caught before it becomes a front-end denial.
Staff currently run eligibility through the payer portal one patient at a time. Deductibles reset at the plan year, coverage changes when a patient switches jobs, and Medicare secondary is usually manual. A check that was clean in the spring can be stale by the fall.
Eligibility errors are a leading cause of front-end PT denials and hit hardest at plan-year resets and employer changes. Availity runs the HIPAA 270/271 eligibility transaction and the 278 prior-auth request; both require a human to log in and interpret results. The AI automates that loop so the front desk handles only the flagged exceptions.
A Relay AI employee watches for cancellations in real time. It texts the first eligible patient on the waitlist, filtered by location, therapist, and authorized visits remaining, and surfaces a one-click confirmation screen for the front desk to approve. It never confirms a booking without human sign-off.
Cancellations hit the calendar all day. The waitlist lives in a spreadsheet or in someone's head. A last-minute cancellation leaves a slot the therapist cannot fill in time, and across locations those empty slots add up to revenue no one recovers.
Each day the AI surfaces utilization by location and flags tomorrow's at-risk slots so the team can fill them before the gap opens. The AI will not book a patient past their authorized visit cap, so a recovered slot never becomes a denied claim. The front desk approves every confirmed booking.
A Relay AI employee monitors the intake queue and detects incomplete fields. It sends targeted follow-up texts and portal nudges to the patient or guarantor. It pulls referral orders from the fax queue, routes them to the correct chart, and flags anything missing for the front-desk lead to resolve in one click. That means the chart is ready before the first appointment, not during it.
New-patient intake goes out by email or portal and much of it comes back incomplete. Staff chase insurance cards, referral orders, and prior records by phone and fax. In PT, a physician referral and plan-of-care order often must be on file before the claim can go out, so a missing item at intake delays billable care and seeds a denial downstream.
Phreesia, IntakeQ/PracticeQ, and SPRY kiosk are the intake tools PT clinics layer on top of the EHR; Relay reads their output rather than replacing them. Missing referral data is a documented WebPT front-office denial driver, and the AI takes that chase off your staff's plate.
A Relay AI employee tracks every provider's re-attestation and enrollment status across all locations. It sends structured reminders ahead of each deadline and queues the renewal for the credentialing coordinator to complete in one session. It alerts the operations lead before a new hire's first scheduled visit, not after a denied claim reveals the gap. Credentialing deadlines stop slipping between locations.
CAQH ProView re-attestation comes due every 120 days. Miss it and payers lose access to the provider's profile, which delays billing. New hires and new locations add payer-enrollment windows that, if missed, mean services delivered under a therapist who cannot yet bill.
Medallion and Verifiable automate enrollment and cross-state license tracking for multi-location therapy groups. The AI employee works on top of CAQH ProView, Medallion, or Verifiable. For a full treatment of credentialing automation across multi-location groups, see the dedicated page on AI credentialing tracking.
Every location is leaking the same revenue.
See where you're leakingWhy layer-on beats rip-and-replace, and beats hiring more staff.
A multi-location PT group in growth mode has three options when back-office workload outpaces the team: migrate to a new all-in-one platform, hire more admin staff at every location, or layer AI employees on top of the stack you already run.
| Approach | What it means in practice | The tradeoff |
|---|---|---|
| Rip-and-replace (new EHR or all-in-one) | 6 to 18 months; staff retraining; data migration risk | Migration on top of a migration, exactly when you have the least slack |
| Hire more staff at each location | Headcount grows linearly with locations; 2 to 6 weeks per hire per location | EHR stays, but workload stays manual |
| Relay AI employees (layer-on) | Builds on top of existing stack; 8 to 12 weeks for the first AI employee build | EHR stays; AI closes the execution gap; same AI employees cover every location |
Rip-and-replace (new EHR or all-in-one)
What it means in practice
6 to 18 months; staff retraining; data migration risk
The tradeoff
Migration on top of a migration, exactly when you have the least slack
Hire more staff at each location
What it means in practice
Headcount grows linearly with locations; 2 to 6 weeks per hire per location
The tradeoff
EHR stays, but workload stays manual
Relay AI employees (layer-on)
What it means in practice
Builds on top of existing stack; 8 to 12 weeks for the first AI employee build
The tradeoff
EHR stays; AI closes the execution gap; same AI employees cover every location
If you are mid-growth and adding locations, a rip-and-replace is a distraction. Relay gives your existing team more reach, on top of the EHR you already trust, without the migration risk. Relay is human-in-the-loop by design, so a staff member at your clinic finalizes every claim and every appeal.
Some PT groups are actively weighing a switch to an AI-native platform that wants to replace their EHR. Relay's position is the opposite: we build on top of what you already run. If you have staff at multiple locations already trained on WebPT, Raintree, or Prompt, displacing that stack has a real cost. Relay closes the back-office execution gaps without requiring it.
For groups that also run ABA or other therapy disciplines and are navigating a platform change in that line of business, see our dedicated page: Thoughtful AI alternative for ABA clinics.
Keep your PT EHR: AI employees that work on top of WebPT, Raintree, Prompt, and more.
Relay is not an EHR and does not want to be one. It reads the data your systems already hold and closes the gap between what the tool stores and what your admin team can act on. Relay layers on top of the stack you already run; we never pitch a rip-and-replace.
Already running WebPT, Raintree, or Prompt? Good. Keep it.
Relay layers on top. In 30 minutes we show you where the back office is leaking.
See where you're leakingKeep WebPT. Relay's AI employees layer on top of the WebPT and Therabill stack, closing the visit-cap, denial, and scheduling gaps WebPT leaves to your staff. WebPT exchanges data over HL7 v2 (ADT, BAR/DFT, MDM/ORU) and works through its partner program rather than a public open API, so the integration runs through that surface. Your WebPT setup stays intact; Relay closes the billing, prior-auth, and denial gaps it leaves to your staff.
Raintree holds authorization, scheduling, and claims in one enterprise system but still needs staff to act on auth expirations, denial patterns, and cross-location reporting gaps. Raintree exposes a RESTful Web API plus FHIR/HL7 and a marketplace (Waystar, Availity, pVerify, Kno2), so Relay reads its data layer and surfaces finalized actions to the billing coordinator. Your team never leaves the platform they know.
Prompt's prior-auth and eligibility visibility exists but lacks AI-driven denial prevention, and auth and scheduling remain partially disconnected. Public open API access is gated to the enterprise tier, so Relay integrates via that path and layers the monitoring and drafting on top. Relay layers on top of the Prompt data layer and closes the prior-auth and denial gaps Prompt leaves to your staff.
Net Health (TherapySource/ReDoc) and Netsmart TheraOffice are common in outpatient and hospital-affiliated rehab. Both treat authorization and compliance reactively rather than proactively. Net Health runs a partner API program; Netsmart exposes a CareConnect FHIR R4 API (scope into TheraOffice should be confirmed with the vendor). Relay layers proactive auth and denial monitoring on top. AI employees for TheraOffice and Net Health automate the prior-auth tracking and denial drafting that those platforms leave to manual staff follow-up.
Fusion's prior-auth tracking is manual and reactive: authorization reports surface expiring approvals but do not auto-initiate renewals or alert front-desk staff in time. Multi-location reporting requires manual reconciliation. No public API is documented, so Relay integrates via scheduled exports or a negotiated vendor data feed and layers proactive auth monitoring and denial drafting on top. AI employees for Fusion PT clinics and Ensora Health practices automate the authorization renewal and denial rework that the platform leaves to your billing team.
Smaller and multi-specialty PT groups often run Tebra, Jane, or SimplePractice. Tebra exposes a SOAP API plus a FHIR R4 endpoint; Jane and SimplePractice have no public API, so Relay relies on sanctioned partner access, scheduled exports, or screen-level automation. On all three the common leak is the same: manual prior auth, no automated denial rework, and weak multi-location reporting. That is exactly what the AI employees take on. On Tebra, Jane, or SimplePractice the principle is the same: close the back-office gap without replacing the tool.
Relay also layers on top of the adjacent tools the EHR does not cover: Availity and pVerify for eligibility and prior auth, Waystar and CollaborateMD for claims and denials, NexHealth and Jane for scheduling, Phreesia and IntakeQ for intake, Updox and Documo for inbound referral fax triage, and Weave, Klara, or Solutionreach for patient outreach and recall. Relay drives the logic; the existing tool stays the channel.
How we build it.
We start from the problem you feel, then build the fix on the systems you already run. Discovery and your first working AI employee take 2 to 3 weeks. The full build runs 8 to 12 weeks.
Start with a free 30-minute call
A short call about where the work is piling up and what that is costing you while it stays manual. No commitment, and you leave knowing where you would start.
Discovery and your first AI employee (weeks 1 to 3)
A few working sessions with your team. We map your operation end to end, every workflow across your locations, and find where the money leaks and what closing it is worth. You do not walk away with just a document. By the end of discovery we have built your first working AI employee on top of the systems you already run, so you see it pay off in your real setup before the full build starts.
The full build (8 to 12 weeks, start to finish)
We build the rest of the AI employees you mapped and wire them across every location. Nothing goes out until your team approves it, so you stay in control the whole way. One pediatric therapy client had all seven locations live within 90 days.
Proof: Sensory Speech & Occupational Therapy.
Sensory Speech & Occupational Therapy is a multi-location pediatric speech and OT therapy group. Relay built two AI employees on top of their existing EHR and Drive, with a staff member finalizing every action.
The intake AI employee ran the full new-client lifecycle: scheduling clinic tours, getting ROIs signed, requesting records and IEPs from schools and prior speech and OT clinics, sending medical orders to the child's PCP and following up until signed, starting authorization renewals about a month out, and sending three-month progress reports and evaluations to PCPs for signature.
The internal auditing AI employee reviewed every note nightly against clinical requirements, confirmed the billing code matched the note, and after billing found and appealed denied claims and reconciled remittances against the EHR notes.
The group saw 100% claim accuracy, staff 33% more productive, claim denials down 12%, and faster documentation turnaround. Every result is attributed to the client's own reporting, and results vary by clinic. The same two-AI-employee pattern, intake plus auditing, applies to multi-location PT groups running the same back-office problems.
Physical therapy AI employees: frequently asked questions.
They read the data your EHR already holds through its API, partner program, or sanctioned export, do the repetitive back-office work, and hand a finalized action to a staff member. Relay is not an EHR and never replaces one. Your team stays on the platform they know.
Yes. The AI employee watches each authorization's remaining visits against the live schedule, flags patients about to hit the cap, and drafts the re-auth packet for staff to submit before the cap is exceeded. The AI never submits to the payer without your sign-off.
It detects cancellations in real time and texts the first eligible patient on the waitlist, filtered by location, therapist, and authorized visits remaining. It surfaces a one-click confirmation for your front desk. Bookings are never confirmed without human sign-off.
No. Relay is human-in-the-loop. The AI drafts the corrected claim or appeal and routes it for review. A billing coordinator approves and submits. The AI never files autonomously.
It reads each denial, groups them by root cause, and drafts the correction or appeal with the note and auth attached. It works the aging queue daily so claims do not age past the filing window unworked.
Yes. The pre-submission check validates payer-specific modifiers, including KX for the Medicare therapy threshold and GP for the PT plan of care. It also verifies time-based codes against the 8-minute rule before a claim is released.
Yes. It runs batch eligibility sweeps against the upcoming schedule, flags lapses, deductible resets, and secondary coverage changes to the front desk before the visit, on top of Availity and your EHR.
The AI reconciles posted ERAs against contracted rates by payer and CPT code. It flags short-paid claims above a threshold for a billing lead to dispute. Underpayment detection runs on the same daily cadence as denial management.
It means AI employees doing the repetitive back-office work on top of the EHR you already run: auth tracking, claim scrubbing, denial drafting, eligibility sweeps. You get the automation without adding another platform to manage or train staff on.
Relay is built for multi-location PT groups (roughly 2 to 20 locations) with insurance-heavy caseloads where the same admin work repeats at every site. That is where AI employees do the most for you.
Yes. Relay operates under a BAA and works within the clinic's existing systems and access controls. A staff member remains the final decision-maker on every action.
The underlying agents do the watching, drafting, and triage; your team finalizes. The goal is to stop admin workload from scaling with every new location, not to remove staff.
The discovery and build phase runs 8 to 12 weeks. The first 2 to 3 weeks are working sessions to map your operation and prioritize the highest-leak workflow. No disruption to current systems during build.
We map where your operation is leaking across prior auth, denials, scheduling, and intake. We show you the specific gaps at each location before any build is discussed.
Stop the leak at every PT location.
Authorization expirations, the denial queue, unfilled slots: these do not fix themselves. They compound with every location you add. In a free 30-minute intro call, we show you exactly where your PT operation is leaking and what an AI employee built for your specific stack would take off your team's plate.