Every session ends in a clean claim.
An ABA clinic with 100+ RBTs doesn't have a note-writing problem — it has a documentation pipeline problem. Notes are typed at night, QA reviews them by hand days later, requirements change faster than the engine, and the QA department has to grow every time the clinic does. The new PraxisNotes rebuilds that pipeline around an AI team: the RBT talks, an agent writes, specialist agents review, and what reaches the biller is ready to submit.
Humans stay on every gate that matters — the RBT signs the note, reviewers handle exceptions, the biller presses submit. The agents do everything in between.
The wall the clinic is hitting
None of these numbers are hypothetical. They're what the industry measures today — and why a QA department that scales with headcount can't win.
Industry data from CentralReach — the market leader's own audit tooling. Rework is the common case, not the exception.
Much of it unpaid, after hours, in a second language. It's the #1 burnout driver in a workforce with 65–103% annual turnover.
~1,000 notes a week at 5–10 minutes of human review each. Grow the clinic, grow the QA payroll. No economy of scale.
HHS-OIG, Colorado 2026: 93 of 100 sampled months failed. Indiana $56M, Maine $45M. Sampling-based QA doesn't survive an audit.
Every payer writes its own note requirements — Florida Medicaid's seven mandatory elements, Optum's stop-clock timestamps, TRICARE's fifteen-item checklist — and they change on the payer's schedule, not yours. In today's PraxisNotes those rules live inside the engine's code, so every update is a software deploy. The new PraxisNotes makes them editable rule packs — updated in minutes, versioned, and enforced by the same agents that write and review every note.
One session, two worlds
Follow a single 4–7pm home session from the moment it ends to the moment it can be billed.
Days of latency against a 2-business-day signing rule. Every handoff is a queue; every queue is payroll.
Signed before the RBT gets home. Humans review exceptions, not everything.
Four moves nobody in ABA has made
We mapped the whole market — CentralReach, Rethink, Motivity, Artemis, the AI-scribe wave, Office Puzzle itself. Everyone generates from typed forms and audits after the fact. These four moves are open, and they compound.
The RBT speaks during the session — Spanish, English, Creole, 90+ languages — and the note comes out as compliant clinical English. Behavioral-health scribes proved voice cuts documentation time by half or more; no ABA platform has it.
When data is missing, today's AI tools quietly make it up — a fabrication risk in a legal record. Our writer agent escalates instead: one tap on the RBT's phone answers "which prompt level did Maria need?" before the note exists.
A BCBA-lens agent checks clinical quality; a QA-lens agent checks every payer rule. Every note, minutes after signing — humans see only exceptions. The QA team stops scaling with headcount and starts scaling with trust.
Florida Medicaid, Optum, TRICARE, Cigna — each payer's note requirements become a versioned, clinic-editable pack that drives both generation and review. A rule change is an edit, not a deploy, and it reaches every agent on their next note.
The window is real but not forever: CentralReach is consolidating, Rethink shipped AI twice in March, Office Puzzle is adding AI drafting. All of them still generate from typed forms and audit after the fact — the agent-native authoring loop is open for roughly 12–24 months. Meet the agents →
What we'll measure
Not vibes — pipeline numbers, instrumented from day one and reviewed together every month.
from session end — beating every payer's signing deadline by default.
compliant-by-construction kills the rework loop at the source.
agents review everything; people review what agents flag.
payer changes a requirement → the clinic edits a pack → every agent complies.
The north-star metric behind all four: first-pass clean-claim rate — every documentation error caught before submission converts a 30–60-day denial loop into a same-day fix. See the phased rollout →
Inside this proposal
Eight pages — the story first, the engineering behind it after. Read in order or jump to what you care about.
An honest audit of the current PraxisNotes and the clinic's pipeline — the form, the engine, the QA wall, the audit exposure.
The RBT's session on mobile, the BCBA's queue, the QA cockpit, the biller's packet — every screen, walked through.
The client's data spine — assessment, plan, per-session measurements, trends over time — and the extension that carries it into Office Puzzle.
Writer, BCBA reviewer, QA compliance, billing prep — their tools, their escalations, the autonomy dial, and how they learn.
The multi-payer rules engine, the note lifecycle that survives an OIG audit, HIPAA/BAA posture, and voice-consent design.
The platform underneath — agents, escalations, files, voice, mobile — what's reused, what's extended, what's genuinely new.
CentralReach, Rethink, Office Puzzle, the AI-scribe wave — what they built, what we adopt, and the lane nobody fills.
A three-week MVP for 1–5 RBTs, the whole org on by week eight, the billing loop at month three — then six months of personalized support.
Why this isn't a two-year build
The new PraxisNotes is a vertical on a platform that already runs in production — Monsoft's Loquent codebase. The agent runtime, the escalation loop that reaches a phone, the draft-review-sign pattern, layered rules, an org file system, streaming multilingual voice, push notifications, and the native mobile shell are shipped, working code. What we build new is exactly the ABA part: the session domain, the note lifecycle, the review pipeline, the payer rule packs, and the Office Puzzle bridge.
That's the difference between assembling on a foundation and pouring one — and it's why current PraxisNotes users get the upgrade too, not a separate product: solo RBTs keep a faster, voice-first version of the app they know, and clinics unlock the team layer on top.
See the reuse map →The RBT talks. The team reviews. The claim goes clean.
That's the whole product. Everything in the next six pages exists to make those three sentences true at 100 RBTs — and at 500.
Start with today's reality →