What We Connect To
Clear Mind Life is an orchestration layer. Each integration handles one job — transcription, eligibility, claims, communications, AI reasoning, payments. We wire them together so you don't have to.
AssemblyAI
Every provider loses 2+ hours a day documenting what they just said. We eliminated that entirely. The Encounter Agent connects to AssemblyAI's real-time streaming WebSocket API during telehealth sessions — transcribing speech in under 300ms, mapping clinical keywords to ICD-10 codes, and constructing a SOAP note before the physician even hangs up.
We open an AssemblyAI WebSocket session at visit start and stream PCM audio chunks. Partial transcripts fire in <300ms.
AssemblyAI labels "Patient" vs "Provider" segments. Our NLP pipeline processes each speaker separately to extract clinical context.
After session end, we pass the full transcript to LeMUR with a clinical SOAP prompt. Output is structured JSON mapped to FHIR Observation resources.
AssemblyAI PII redaction fires before any text touches our FHIR pipeline — zero PHI leaks into the LLM layer.
NexHealth
The moment a patient books an appointment, a race begins: verify insurance, confirm demographics, send intake forms, collect payment. Most practices lose this race. By wiring the NexHealth API to our Receptionist Agent, we win it automatically — hours before the patient arrives.
NexHealth fires a webhook the moment an appointment is created. Our agent immediately kicks off an insurance eligibility check via the X12 270 EDI pipeline.
We pull the patient's demographics, insurance info, and appointment history via the NexHealth REST API and store them as FHIR Patient and Coverage resources.
We use the NexHealth Forms API to push AI-generated intake questionnaires directly to the patient's phone — pre-filled with existing demographics, requiring only signature.
After insurance is verified, we calculate the exact copay via X12 271 and push a Stripe payment link into the NexHealth message thread — cash collected before the visit.
Twilio
Denied claims don't have to sit in a queue for 90 days. Our Billing Agent uses Twilio Programmable Voice to automatically call payers, navigate IVR trees, and document denial reasons — then uses Twilio SMS to keep patients updated on their balance and payment status in real-time. No human billers on hold.
When a claim is flagged for denial risk, the agent initiates a Twilio outbound call to the payer's provider services line. TwiML scripts navigate IVR trees and capture denial codes via DTMF and speech detection.
After each claim milestone (submitted, approved, underpaid), we send a Twilio SMS to the patient with their balance, EOB summary, and a secure payment link — no confusing paper bills.
For international or preferred patients, we deliver the Explanation of Benefits as a WhatsApp message via the Twilio Business API — same data, frictionless UX.
Every call outcome (IVR navigation result, denial code captured) fires a webhook back to our FastAPI backend, triggering the next step in the appeals workflow automatically.
Availity
Insurance verification is the first place revenue leaks. We connect to Availity's real-time eligibility API — the same network used by 2,000+ payers — to run X12 270 requests the moment an appointment is booked. The X12 271 response comes back in seconds: copay, deductible, network status, prior auth requirements.
Sent automatically when NexHealth fires an appointment webhook. We query the patient's insurance plan, member ID, and date of service.
We parse the 271 response into structured data: copay amount, deductible met/remaining, in-network status, and any prior auth requirements.
If the 271 response indicates prior auth is required, we flag the appointment and trigger the auth workflow — 48 hours before the visit.
Once copay is confirmed, we push a Stripe payment link to the patient via Twilio SMS. Cash collected before the visit.
OpenAI
The transcription is done. The SOAP note is structured. Now the hard part: which ICD-10 code is correct? Which CPT code matches the complexity level? We use GPT-4o-mini with a clinical prompt — trained on our 74,719-code ICD-10 database — to suggest codes with confidence scores. The provider reviews, not re-codes.
We pass the de-identified SOAP note to GPT-4o-mini with a structured prompt. It returns ranked ICD-10 suggestions with confidence scores and rationale.
The model analyzes medical decision-making complexity (MDM) to suggest the appropriate E&M level (99213 vs 99214 vs 99215).
Before claim submission, we run the code combination through the model to predict denial probability based on payer patterns.
All patient identifiers are removed before the text reaches the OpenAI API. The model sees clinical content only — no names, DOBs, or member IDs.
Stripe
Collecting copays at checkout is the worst time to collect them. Patients are leaving, staff are busy, and the conversation is awkward. We collect copays at booking — after eligibility confirms the amount — via a Stripe payment link sent by SMS. By the time the patient arrives, the copay is already paid.
After Availity confirms the copay amount, we generate a Stripe Payment Link and send it via Twilio SMS. Patients pay on their phone before the visit.
When the 835 remittance comes back, we calculate the patient balance and send a Stripe invoice automatically. No paper bills.
For group practices, we use Stripe Connect to route payments to the correct provider account automatically.
If a claim is overpaid or a copay was collected incorrectly, we trigger a Stripe refund automatically when the 835 remittance indicates an overpayment.
How They Work Together
Each API handles one layer of the revenue cycle. Our orchestration layer connects them into a single, autonomous pipeline.