AI answers as a compliance surface
AI assistants make claims about your institution to the public: about insurance, eligibility, rates, and legitimacy. Some are wrong, and none are currently in your review process. Scanley measures what is being said, documents it with method and dates, and routes every correction through your approval before anything ships.
Every finding states which engine, which day, how many runs, and what was said. The method statement travels on the report itself, written for a reviewer who checks line by line.
Fixes are drafts until a named person at your institution approves them. The record keeps who approved what, when.
Public web only. No member PII, no system access, no credentials. Due-diligence materials arrive before you ask.
A missing measurement is never presented as a miss. That rule is enforced in the software, not the copy.
What happens when an answer misstates share insurance?
It is flagged as an accuracy finding, documented with the engine, date, and exact text, and becomes a drafted correction routed through your approval process. Misstatements about insurance or eligibility are treated as the compliance findings they are, not as marketing gaps.
Can we see the methodology before engaging?
It is on every report, publicly, including the sample. Engines measured, runs, dates, what counts in each denominator, and what was deliberately not measured. Nothing about the method is proprietary; the work is in the doing.