Throughline

For behavioral & mental health practices

Throughline gives your evenings back — and you still sign off on every note.

You went into this work to do therapy, not to chase denials at 9pm. Throughline sits beside you — solo clinician or 10-person group — maps how your practice actually runs on top of PHI and Part 2 records, and builds the AI that takes the busywork off your evenings. Done with you, from $499/mo. You approve everything before it touches a client.

8 min read

See pricing

The shift

In behavioral health the constraint isn't demand — it's hours. Roughly 169 million Americans now live in a Mental Health Professional Shortage Area, about 6 in 10 psychologists report no openings for new patients, and the national wait runs near 48 days. Your clinical hour — the one thing only you can do — gets sandwiched by the lowest-value parts of the day: scheduling friction, intake paperwork, and claim-chasing. AI's real job here isn't to replace the clinician; it's to give the clinician's hours back.

Your reality

Your revenue is sessions delivered times rate times what you actually collect — and all three leak. No-shows run 20–30% in therapy and up to 50% in SUD programs. About 30% of mental-health claims were denied in 2023 versus 19% of all other claims, yet most appealed mental-health denials get overturned when someone has time to file them — and most owners don't. Documentation eats 25–50% of work hours, and 60–70% of clinicians finish notes after hours. The margin looks high until you notice there's no cushion in it: income scales with your own clinical hours, and a missed session is pure, unrecoverable loss.

~30%

of mental-health claims denied in 2023, vs. 19% of all other claims

~48 days

the national wait for a first appointment

60-70%

of clinicians finish their notes after hours

~5 hrs/wk

saved by early AI-note adopters in 2025

Builds inside the EHR you already run

SimplePracticeTherapyNotesTheraNestValantKareoAdvancedMD

Sound familiar?

You blocked 10 minutes between sessions for notes. It went to the next client. Now the SOAP and DAP notes are stacked for tonight — pajama time, and its own low-grade dread.
The denials pile up faster than you can work them. You half-suspect most would win on appeal, but appeals take time you spend in session, so they quietly become write-offs.
A prospective client called during your 2pm. By the time you called back, they'd moved on. There's a waitlist you can't clear and revenue going to voicemail, at the same time.
Nobody taught you the business. You learned billing by getting denied and scheduling by getting stood up — and now everyone's telling you to just add one more tool.
What scares you most isn't cost — it's a mistake in the chart. An AI scribe that mis-transcribes a session, or records a client without telling them, is a licensing-board complaint and, for SUD records, a Part 2 problem.

Where AI fits

Finish your notes before you leave the office.

A behavioral-health-trained scribe drafts the progress note in your format — SOAP, DAP, or BIRP — and pre-links it to the treatment plan so the golden thread of medical necessity stays intact, ambient during the session or from your post-session dictation.

The scribe captures the session, drafts the note in your note type, and stages it in the chart pre-linked to the treatment plan and diagnosis for your review.

Notes done on time
Less after-hours documentation
The golden thread stays connected for reimbursement

Watch for: You review, edit, and sign every note before it enters the chart — non-negotiable after 2025 lawsuits over erroneous AI chart entries. · Patient consent to recording is captured per your state's two-party-consent law. · For SUD records the scribe runs under a Part 2-compliant BAA that forbids training on your sessions.

Best for: Solo clinicians and groups drowning in pajama-time documentation.

Stop re-keying insurance cards into the EHR.

Bad intake data is what triggers eligibility denials downstream. AI reads the insurance card, ID, and intake and consent forms, drops the structured fields into the right EHR slots, and flags anything that doesn't match.

Uploaded or scanned intake documents are extracted and classified, demographics and insurance route into your EHR fields, and mismatches surface for exception review.

Less manual keying
Fewer eligibility-driven denials
Front-desk hours back

Watch for: Your coordinator confirms extracted insurance and demographic fields before they commit. · Manual entry becomes exception review, not blind trust. · Mismatches are flagged for a person rather than silently saved.

Best for: Group practices and front desks buried in intake paperwork.

Stop writing off denials you'd win.

Around 30% of mental-health claims were denied in 2023 versus 19% of all other claims — but most appealed mental-health denials get overturned. The gap is time. AI reads your ERAs and EOBs, sorts denials by reason, and drafts the appeal so the cost of fighting back drops to a review-and-send.

AI parses remittances, classifies the denial (missing auth, medical necessity, coding, eligibility), and drafts a parity-grounded appeal letter with the relevant chart citations pulled in.

Denials triaged, not buried
Appeals drafted in minutes
Revenue recovered instead of written off

Watch for: Your biller approves and submits every appeal. · AI drafts; it never auto-files to a payer. · Each appeal cites your record, not boilerplate.

Best for: Insurance-based practices losing real money to the denial queue.

Convert the waitlist instead of losing it to voicemail.

A prospective client shouldn't fall off because you were in session when they called. A guided flow collects consent and insurance, verifies eligibility, and books the first open slot — with you in the loop on clinical fit before anything is confirmed.

The intake flow gathers consent and insurance, runs eligibility verification, proposes the first available slot, and sends the welcome packet once you approve the match.

Fewer first-appointment no-shows
A waitlist that actually clears
No more revenue lost to voicemail

Watch for: You or your coordinator approve clinical fit — right modality, specialty, acuity — before booking. · Any high-acuity or safety-flagged intake routes straight to a human, never to automation. · Nothing is confirmed until you sign off on the match.

Best for: Practices with a waitlist they can't clear and a phone they can't always answer.

Answer the routine questions you can't pick up mid-session.

A policy-grounded assistant handles the questions that don't need you — hours, location, telehealth setup, which insurances you take, your cancellation policy, billing — and drafts replies to portal messages for your review.

The assistant answers from your own policies and FAQs, drafts portal-message replies, and hands anything outside its lane to a person.

Fewer missed calls
Faster replies to routine messages
Your phone stops competing with your sessions

Watch for: Anything clinical bypasses AI entirely and surfaces a human. · Any message mentioning self-harm or crisis routes to a person plus a crisis-resource path immediately. · This is the sharpest line in the whole map, and we draw it absolutely.

Best for: Solo owners who can't answer the phone during a session, and groups with overflow on the front desk.

Build the auth packet from the chart you already wrote.

Mental-health services need prior authorization 5.4x more often than comparable medical care, and coordinators assemble those packets by hand. An agent assembles the request from the chart and stages it for one human approval.

The agent pulls diagnosis, CPT, treatment plan, and a medical-necessity narrative from the chart, assembles the auth request, and stages it for review.

Auth packets assembled, not hand-built
Less time in the UR queue
The medical-necessity narrative stays grounded in the record

Watch for: Your coordinator reviews and submits. · AI assembles the authorization; it never submits it. · The narrative is pulled from the chart, not invented.

Best for: Group and psychiatric practices where prior auth and utilization review eat coordinator hours.

Safe to start vs. proceed with guardrails

Safe to start now

Drafting progress notes for your review and signature — the note is a draft until you sign it.
Pre-filling EHR fields from intake documents, with your coordinator confirming before commit.
Triaging denials and drafting appeals for a biller to approve and send.
Answering routine, non-clinical FAQs — hours, insurance, telehealth setup, cancellation policy.
Smart appointment reminders and confirmation nudges to attack no-shows.
Grounded, cited internal search across your policies and payer rules.

Proceed with guardrails

Anything that records a session — requires patient consent built into the workflow per your state's two-party-consent recording law.
Any vendor that touches PHI — requires a signed BAA, and one that contractually forbids training their models on your sessions.
Anything touching SUD records — must honor 42 CFR Part 2 consent and redisclosure rules, not just HIPAA.
Any patient-facing clinical, crisis, or safety message — must bypass AI and route to a human plus crisis resources.
Any note entering the chart — must carry a clinician's review and signature; an erroneous AI entry is a board complaint.
Any patient-facing cost document — a Good Faith Estimate or superbill — the owner reviews before it goes out.

Why do it with us

Hire a consultant	High hourly rate, a long discovery phase, and a deck at the end	By the time it's actionable, you're still doing notes at 9pm
Hire an engineer	A salary your margins can't absorb	You'd have to teach them medical necessity, the golden thread, and Part 2 first
DIY on nights and weekends	You're already therapist, biller, marketer, and compliance officer	You don't need a sixth job; you need your evenings back
systemlevel.ai — done with you, from $499/mo	One senior expert scopes the work against HIPAA and Part 2 and builds inside your EHR	You approve everything, and there's no long-term contract

AI note adoption among independent clinicians grew from near 0% to about 10.2% in 2025, with early adopters reporting roughly 5 hours a week saved (SimplePractice, 2025). The trend is real and early.
The most rigorous study to date — 1,800 clinicians across five academic centers — found ambient AI scribes saved only about 16 minutes of documentation per eight patient-hours and no significant cut to after-hours time, yet were still tied to large burnout improvement (STAT, 2026). The durable win is finished-on-time notes and less dread, not a dramatic minute-count, and we'll tell you that up front.
A vendor-reported five-clinician practice claimed 27-plus hours a week of documentation saved (Mentalyc). Treat that as marketing, not gospel — which is exactly why we measure your numbers, not theirs.
On denials, about 30% of mental-health claims were denied in 2023 versus 19% of all other claims, and most appealed mental-health denials get overturned. The lost revenue is recoverable if someone has time to appeal — and most owners don't.
Our rule of thumb: if an error would end up in front of a licensing board or a client in crisis, a human signs off before it ships. No exceptions.

Questions you’re probably asking

Isn't AI listening to a therapy session a privacy violation?: It can be — there were 2025 lawsuits over patients not being told AI was recording. So Throughline only recommends vendors under a signed BAA that contractually forbids training their models on your sessions, builds the patient-consent step into the workflow per your state's recording law, and for SUD practices honors Part 2 redisclosure rules. You approve every note before it enters the chart.
My SUD records are under 42 CFR Part 2 — most tools don't even know what that is.: We do, and we scope every workflow against Part 2's consent and redisclosure requirements — including the 2024 final rule's breach-notification alignment — before anything goes live, not just HIPAA. If a vendor can't meet Part 2, it doesn't get recommended for your SUD records.
An AI mistake in a clinical note is a licensing-board complaint waiting to happen.: Agreed — which is why the note is a draft until you sign it. In Throughline the human checkpoint is the product, not an add-on. The golden thread stays yours.
I've heard the time savings are overhyped.: Some of it is — the rigorous data shows modest clock-minutes. So we don't sell minutes. We measure your real numbers — after-hours notes finished, no-show rate, denials recovered — and start with the workflow that has the clearest payoff, not the flashiest demo.
I don't have time to learn another tool.: That's the point of done-with-you. Throughline maps your existing EHR — SimplePractice, TherapyNotes, and the rest — builds inside it, and trains your team, from $499/mo with no long-term contract. You don't take on a sixth job; we sit beside the one you already have.
Where do I even start?: We rank the AI candidates by value and risk and start with one workflow — usually notes or intake — then measure it before we touch anything else.

Pricing

Operator — $499/mo

Monthly strategy, your first workflows mapped and scoped (usually notes or intake), vendor and model recommendations vetted for BAA and Part 2 fit, a read on your current EHR stack, and email support.

Best for: The solo owner-clinician who needs a clear first move.

Practice — $999/mo · Flagship

Everything in Operator, plus bi-weekly working sessions, implementation guidance through go-live, architecture and integration review against HIPAA and Part 2, team training with prompt libraries for your clinicians and front desk, and a direct line for unblocking.

Best for: The group practice ready to ship.