AI Eligibility Suggestions
On every claim, AI surfaces structured eligibility flags with rule citations and missing-information pointers. Staff decide; AI starts the draft.
Key benefits
- · Recommends the eligibility decision (approve / deny / more-info-needed) with reasoning
- · Flags missing information required to decide
- · Explains applicable eligibility rules for the specific case, with citations
- · Per-benefit-category guidance (medical yes, lost-wages no, etc.)
What it does
When a staff reviewer opens a claim (VcClaim) or benefit (ExpenseBill, IncomeLoss, TreatmentPlan), VCPMS runs a background LLM analysis that checks the claim against the program’s knowledge base of eligibility rules. The result surfaces as a structured list of flags — critical, warning, or info — with each flag citing the rule it references and pointing to the data point on the claim.
What “AI suggestions” actually means here
Not a free-form chatbot. The assessment is:
- Structured. Output is parsed as JSON with severity levels (critical / warning / info), cited rules, and data-point references.
- Cached. Results are cached by claim + aggregate-modified-datetime so repeated page views don’t re-call the LLM.
- Grounded. Rules come from the tenant’s configured knowledge-base articles — not the model’s general knowledge.
- Advisory. The decision is always the reviewer’s. The suggestion becomes part of the record; the decision is made by a person.
Staff workflow
Open a claim → review auto-surfaced flags → resolve each one (accept the suggestion, override with reason, or request more info from the claimant via SecMail or Dynamic Form) → make the eligibility decision → the decision is captured, along with the flags that were resolved, on the Award snapshot.
Supplementary free-form Q&A
Reviewers can also ask free-form questions about the specific claim — “what previous claims did this claimant file?” or “summarize the injury pattern” — with answers grounded in the claim record.