Sentinel
The autonomous auditor for ABA clinical documentation.
Payer audits, denials, and clawbacks start with documentation. Sentinel is Coralia's answer: an AI system that reviews session notes when clinicians send them for review, flags compliance and quality problems against a clinical rule catalog, drafts the corrections, and leaves the decision to your team.
How Sentinel works
1.Reads each note sent for review
When a clinician sends a note for review before submitting it — RBT direct-service notes, BCBA protocol-modification notes, caregiver-guidance notes — Sentinel reads the whole note. Nothing in it is sampled or skipped.
2.Checks against a clinical catalog
Each note is evaluated against an ABA-specific catalog of documentation controls: prohibited clinical language with proper replacements, required narrative elements, and payer-facing quality rules. Every finding cites the specific rule it violates.
3.Drafts the correction
Sentinel rewrites the narrative resolving every finding — in the therapist's own voice, changing nothing that was clinically stated, inventing nothing that wasn't.
4.Read twice, independently
Each part of the catalog is audited by two independent passes, and a third step rules on whatever the two readings disagree about. The correction must also pass the documentation rules, checked in code, before it is accepted.
5.People stay in control
Sentinel never edits a clinical record itself. The author decides what to take from the correction and signs the note, and your analyst reviews and approves it — autonomy in the analysis, humans in control of the record.
What a Sentinel finding looks like
A worked example — the note and child are fictional; the mechanics are what your team reviews.
The narrative as written
“Alex was aggressive today and refused to work for most of the session. Toward the end he calmed down and we finished the token program.”
Sentinel flags — each citing its rule
Subjective label, no observable behavior. “Was aggressive” names a category, not what happened — audit reviewers hold narratives to observable, measurable description.
Outcome with no data behind it. “Most of the session” and “finished the token program” give an auditor nothing to match against the billed time or the targeted goals.
The correction Sentinel drafts
“During the first activity, Alex pushed materials off the table twice and left his seat three times when task demands were presented. Staff followed the planned redirection procedure. In the final twenty minutes, Alex returned to his seat within ten seconds of each prompt and completed eight of ten token-program trials.”
Same session, same voice, nothing invented — the observable events the original glossed over, stated so the note supports the claim. The correction is checked against the documentation rules, and a person decides what goes into the record.
Why this is different from AI note writers
Most ABA software now offers AI that helps you draft a note faster. Sentinel solves the harder problem: whether the notes your team already wrote would survive scrutiny. A drafting assistant saves minutes; an auditor protects the revenue behind every session and the clinical record behind every child.
That distinction matters because payers don't audit your drafting speed — they audit your documentation. Cloned notes, vague narratives, and missing elements are what trigger recoupments.
Documentation problems surface while the note is still a draft — not in a payer audit two years later.
Prohibited or risky clinical language is caught before a note supports a claim.
Every note type is held to the same standard, whoever wrote it, however busy the week was.
Every finding cites the catalog rule it comes from, so reviewers can see why it was raised.
Documentation that's ready before anyone asks.
Sentinel ships as part of Coralia — the AI-first, all-in-one platform for ABA agencies.
Want the documentation rules themselves? Read the session-note guide →