AYAANIS
Choose your portal
VeridenceFacilitiesRXLensPharmaciesEnterpriseMulti-locationNot a client yet? Request access →
AI is arriving in skilled nursing, and the people who defend MDS audits for a living are drawing a line. Here is where it sits and how to use AI on the right side of it.
Published 2026-10-15 · AYAANIS Technologies, Philadelphia
For information only. This guide is general background for operators and their teams. It is not legal, regulatory or financial advice, and rules change by state and by payer. Check the current source documents linked at the bottom, and your own contracts, before acting.
A growing category of software reads the clinical record and suggests MDS answers while the assessment is open[1]. The pitch is speed and reimbursement. The concern, from MDS and reimbursement consultants, is what happens at audit: "You will be open to compliance issues and audit risks if you code based on AI-provided answers"[2].
Zimmet Healthcare's guidance for operators is direct: validate any AI-assisted MDS workflow with clinical oversight, and avoid auto-suggestion tools because they create compliance risk[3]. Their reasoning is that CMS has not issued guidance on AI-enabled MDS tools, but the existing standard is clear: the MDS requires clinical judgment, interdisciplinary collaboration and RN validation. "The MDS must reflect direct clinical input, not AI probability matrices"[3].
The distinction is between AI that writes into the assessment and AI that reads the record and reports what is already there. The first changes the clinical claim; the second checks it.
Reading is safe ground. An AI that says "this resident fell on the 3rd, the incident note exists, the care plan was not updated" has not made a clinical decision. It has pointed at two documents and the gap between them. The nurse decides what to do. That is also exactly what a surveyor does, which is why read-only review maps so well to survey readiness.
| Use | What the AI does | Who owns the clinical claim | Audit posture |
|---|---|---|---|
| Auto-suggest MDS coding | Proposes answers inside the assessment | Unclear; the suggestion shapes the answer | Weak. Experts advise against it |
| Ambient scribe | Drafts notes from conversation | The clinician who signs | Moderate; depends on review |
| Read-only record review | Reads existing charts, flags gaps and evidence | The facility; nothing changes | Strong. The record is the record |
Veridence reads the facility's record every morning and ranks what a surveyor would cite: the resident, the F-tag, the evidence. It never writes to the chart it reads and it does not suggest MDS answers. Every number on screen opens to the source document, so the claim is provable from the chart, not from the software.
The same design serves audit defense. When an MAC, RAC or state auditor asks why a claim was coded, the answer is in the record, where it always was. The software's job was to notice when it was not.
CMS has not issued specific guidance on AI-enabled MDS tools. The existing standards still apply: clinical judgment, interdisciplinary input and RN validation. Industry guidance recommends clinical oversight of any AI-assisted workflow and avoiding auto-suggestion tools[3].
Read-only AI reads the existing record and reports gaps and evidence; nothing in the chart changes. Auto-suggest AI proposes answers inside the assessment, which then shape the clinical claim submitted for payment.
Indirectly. It catches documentation that would fail an audit or a survey, and it flags changes in condition that the team has not yet assessed. The coding decision stays with the MDS nurse.
Zimmet recommends amending the compliance program to reflect any AI embedded in the MDS workflow, with policies for unbiased authentication of assessments[3].
Veridence runs read-only over your EHR, names the F-tag behind every finding, and opens the evidence. See it over your own building.