What it takes off your week
The fix itself is clinical work and it is yours. What eats the days afterwards is assembly: reading back through notes and orders for the evidence, writing four parts for every tag, keeping the dates and the monitoring consistent across all of them, and setting the whole thing out in the federal format.
That assembly is what arrives already drafted. The work changes from writing to reviewing — and on a survey carrying a dozen tags, that is the difference that puts your DON back in in-services, back on the audits that prove the fix held, and back on the floor.
What actually happens
You upload the 2567 as you received it — the scanned PDF is fine. It is read tag by tag: the F-tag, the scope and severity letter, and the surveyor's own findings for that tag.
Each tag then comes back as a draft with the four things a plan of correction has to carry: how the affected residents were corrected, how you will identify others potentially affected, what systemic change prevents recurrence, and how it will be monitored — with dates, and with a named role against each one.
Where the chart supports a statement, the draft quotes the note or order it relied on, with its date. You are editing something assembled from your own record, not approving prose written around a gap.
Why we need your 2567, and cannot work from a public lookup
CMS publishes every citation: the facility, the date, the F-tag, the scope and severity. What it does not publish is the surveyor's narrative. The description in the public file is the standard regulatory text for that tag — identical in every state, for every facility cited under it, in every year.
So a draft built from public data could only ever restate the regulation back at you. The specifics a surveyor wrote about your building exist in one place: the document they handed you. Anyone offering to draft your plan of correction without it is drafting against the regulation, not against your findings.
It reads more like your building every survey
A plan that was accepted is the best possible guide to the next one. In most buildings that knowledge lives in one person's memory, or in a folder nobody opens until the week they need it.
So when a plan comes back accepted, it goes into a library with its outcome. The next draft for the same F-tag in your state is written with those accepted plans in front of it — used as a guide to structure, specificity and length, never copied. Where something was rejected before, the reasons are carried forward as weaknesses to take particular care with.
It picks up how your building writes, too: how long your sections run, how formal the tone, which roles you name by title. The second survey's draft sounds more like you than the first one did, and the fifth more again.
CMS sets what it must contain. Your state decides whether to accept it.
Worth being exact about, because the distinction is the whole thing when a plan comes back rejected. CMS defines the elements a plan of correction has to carry and the format it is filed in. The state survey agency is who reads yours and accepts or rejects it, and states differ in what they will take.
So no vendor's wording is "CMS-approved" — ours included — and anyone telling you otherwise is selling something. What we can do is follow the structure CMS requires and learn from the plans your own state agency actually accepted, which is the nearer thing to what you wanted anyway.
A quote has to survive being checked
Generating confident corrective-action language is easy, and that is the problem. The useful question is whether the evidence behind a sentence is real.
Every quotation in a draft is verified against the records it was drawn from. If a quote cannot be matched back to a note, an order or an incident we actually hold, it does not ship in the draft.
That rule has a consequence worth stating plainly, because it is the kind of thing that is usually hidden. We index your assessments — types, dates and status — but we do not hold the answer bodies. So assessments are deliberately kept outside what can be quoted, and a draft that tried to cite one as evidence of what an assessment said fails that check rather than slipping through.
What the draft will not do
- It does not sign. The signature block is yours and nothing writes into it.
- It does not practise medicine. It surfaces documentation; clinical decisions stay with your clinicians.
- It does not invent a date. Completion dates are proposed against your survey date and are editable; nothing is filed on your behalf.
- It does not hide what it could not find. A tag where the chart gave it nothing to work from says so, rather than producing fluent text with no evidence underneath.
Out the other end
Your DON edits it, then signs it. It exports in the federal format, with the audit trail attached — who changed what, and when. The export carries a disclosure that parts of the draft were AI-assisted, which is both the honest position and, from November 2026, the required one.
Completion dates then come back as reminders to the people who own them, because a plan of correction is a set of promises with dates on it and the dates are the part that gets lost.
The 2567 carries resident identifiers
It is a clinical document and it names people. So we do not ask you to email it. The upload happens inside the product, under a Business Associate Agreement — a one-page document, signed before anything is uploaded.
If you would rather see the thing work before signing anything at all, that is reasonable, and it is what the walkthrough is for: pick any tag and we will draft it together against a worked example, with nothing of yours sent anywhere.
Related: What a citation costs your rating · After an incident · Security and your BAA