Pre-auth packets in 3 minutes, not 90.
AI drafts that read like a clinician wrote them — not a chatbot. Your BCBAs review and submit. Compliance enforced server-side; auditor gets a clean CSV.
Cairn cites only facts from your treatment plan and assessment. No invented goals. No fabricated scores. No history that wasn't already in your records.
The client, a 6-year-old individual with a primary diagnosis of Autism Spectrum Disorder (F84.0), requires Applied Behavior Analysis services in a home-based setting to address significant behavioral and skill deficits across multiple developmental domains. The most recent VB-MAPP, administered on April 1, indicates substantial gaps in mand frequency, tact repertoire, and listener responding consistent with the client's current level of functioning. Treatment goals target measurable behaviors drawn directly from the authorizing assessment: vocal manding for preferred items (baseline 20% across 5 sessions, mastery criterion 80% across 3 consecutive sessions), labeling common household objects (5/20 items at baseline), and intraverbal exchanges. Discharge criteria are tied to demonstrated generalization across settings and consistent independent performance over the defined criterion period.
Real output from Cairn's Magellan template. 1,247 chars total — card truncates ~600.
What Cairn does
Drafts that read like a clinician wrote them, not a chatbot.
The AI receives your assessment scores, treatment plan goals, and the payer's specific medical-necessity criteria. It cites VB-MAPP / ABLLS-R / AFLS scores verbatim and quotes goal targets from the plan — never invented.
Refuses to fabricate. We test for refusal-tokens and unfilled placeholders before persisting any draft.
Payer-specific guidance built in.
Pre-loaded with what each reviewer expects: Magellan wants discharge criteria; BCBS wants evidence-based protocols (DTT, NET, PRT); Cigna wants ADL outcomes; Medicaid MCOs want less-restrictive-alternatives analysis.
Magellan, BCBS / Anthem, Cigna, Aetna, Medicaid MCO templates ship out of the box.
Wedge alongside what you have. No migration.
Keep CentralReach, Rethink, your homegrown spreadsheets — whatever. Cairn handles one workflow really well: pre-auth packets in, service authorizations out. Your other tools don't change.
No CSV import wizard. No data migration call. Sign up and draft a packet in five minutes.
How it works
From the BCBA's seat. Total time per packet: 5–10 minutes — vs. 60–90 by hand.
Set up your org
Sign up with email + password. Cairn enables the three ABA CPT codes by default (97153 / 97155 / 97156). Add clients as you need them — we don't require an upfront import.
Compose a packet
Pick the client → treatment plan → payer → service codes → period → hours/wk. The AI drafts the narrative in 2-3 seconds against the payer-specific template.
Review and submit
Edit the draft if it needs it. Submit creates pending service_authorizations. When the payer responds, mark approved — auths flip active, your RBTs can clock in against them.
What happens if the AI fails on a packet you need today? Cairn shows the failure (rate-limited / model down / prompt unsafe) and you write the narrative manually — same as you do today. The PWA never blocks submission on AI availability.
Questions BCBAs ask first
How is this different from telling ChatGPT to draft a pre-auth packet?
ChatGPT doesn't know your client's VB-MAPP scores, your treatment plan goals, or the payer's medical-necessity criteria. Cairn does — it pulls from your records and the payer template before drafting. Generic AI gets denials; payer-aware AI gets approvals.
Do we have to integrate with CentralReach (or Rethink, etc.)?
No. Cairn doesn't touch your existing platform. You enter the assessment and treatment plan in Cairn (or copy-paste from CentralReach), draft the packet, submit it. Your existing platform keeps doing whatever it does.
What happens to our data? HIPAA?
We sign BAAs. We don't train AI models on your data. The AI calls are stateless — Groq / Anthropic don't retain prompts or responses past the request. Hosted on a single-tenant Postgres in US-East with daily encrypted backups.
What if the AI hallucinates a goal we didn't have?
Two guards. (1) The prompt explicitly forbids fabrication — 'cite only facts in the input.' (2) The BCBA reviews and edits before submit — the AI's draft is never the final document. We track edit-distance metrics so we can see when the AI drifts off.
Pricing
Pilot — through end of 2026
Free
First 5 customers. Cairn is at 0/5 today.
After pilot
$99 / BCBA / month
Flat. No per-packet fees. No migration cost. Cancel any time.
The pilot trade: a 20-minute monthly check-in so we can iterate on what your team actually needs. We'll add payer templates you bill against, smooth out the workflow rough edges, and measure narrative-quality together.
See a packet drafted for your clinic.
15 minutes on a call. We use real anonymized data from a clinic like yours. If it's not better than what your BCBAs ship today, we both leave smarter.
We sign BAAs. We don't train AI models on your data. Hosted on a single-tenant Postgres in US-East with daily encrypted backups in Cloudflare R2.