Product · In pilot
Plancite
Plancite answers plan-level benefit and precertification questions from approved, effective-dated documents. Rule code makes every decision, and every answer links to its source section and page.
Overview
What it's for
Built for: Teams that answer benefit and precertification questions
Benefit and precertification questions are usually answered by searching long plan documents and precertification lists by hand. It's slow. Two people can reach different answers, and it's hard to show later why an answer was given.
Plancite replaces the search with a decision you can check. It picks the approved document for the group, plan, line of business and service date before it reads any text. Rule code then finds the governing provision and reports coverage and precertification separately. When an input is missing, it asks for it instead of guessing.
Answers that need a person go to a work queue. Every determination is stored with its full trace and can be replayed, including what Plancite would have answered for a past service date.
Process
How it works
Each step is marked with who handles it: rule code, the AI layer, or a person.
Ask
AIFill in a form, or ask a plain-English question that the AI turns into the same fields.
Confirm
PersonAny field the AI inferred is shown for you to confirm before anything runs.
Check inputs
Rule codeGroup, plan, procedure code, place of service, network status and service date. Missing items are requested, not guessed.
Select the document
Rule codeBy group, plan, document type, approval status and effective date.
Apply precedence
Rule codeSeven levels, from code-specific provisions down to general rules. Disagreement at the same level becomes a conflict.
Check precertification
Rule codeReported apart from coverage, with notification, clinical review and medical-policy review as their own fields.
Record the trace
Rule codeEvery step is recorded with its sources and page numbers.
Explain
AIAn explanation is written from the decision, then checked against it before anyone sees it.
Route
PersonAnything Plancite can't determine opens a work item for an analyst to resolve.
Capabilities
What it does
Decisions made by rule code
Tested rule code decides coverage and precertification, not an AI model. The same inputs always give the same answer.
A full trace on every answer
Each answer lists its inputs, the document and version used, the rules matched, the precedence steps applied, any conflicts, and the source section and page.
Right document, right date
Documents and rules are versioned by effective date. Plancite selects by group, plan, line of business, service date and approval status before any text search.
AI that explains but never decides
You can ask in plain English. The AI layer turns the question into fields for you to confirm, then explains the result. A validator rejects any explanation that changes the status, invents a number or cites a source the decision didn't use.
Documents and rules under approval
Nothing extracted from a document counts until a person approves it. Changes are proposed, approved and then published, and the person who proposes a change can't publish it.
Conflicts shown, not hidden
When two provisions at the same level disagree, the answer is Conflict, with both sources attached. Plancite never quietly picks one.
Change impact before publishing
Before a precertification rule change goes live, reviewers can see which past requests and test scenarios would get a different answer. Earlier service dates keep their original answer.
Audit and replay
Every determination is recorded and can be replayed. An append-only activity log shows who reviewed, approved, published or changed what.
Guardrails
Rules built in
Plancite applies these to every answer.
No precertification requirement does not prove coverage.
A diagnosis alone does not prove coverage.
A broad category does not establish code-level coverage.
A code that isn't listed is neither covered nor excluded.
Outpatient does not mean physician office.
An authorization is never a guarantee of payment.
Audience
Who it's for
Benefits analysts and service staff who answer coverage and precertification questions.
Utilization-management and precertification teams.
Benefit configuration and document teams who maintain plan documents and precertification lists.
Compliance, audit and quality teams who need to know why an answer was given, and what it would have been on a past date.
Deployment
How it runs and where it stands
Deployment
Runs on your own server in Docker containers. Documents stay on that server.
Staff sign in through your identity provider (OpenID Connect), and its groups map to built-in roles such as analyst, reviewer, publisher and auditor.
Works with no external AI service and no network access. Real documents are never sent to an outside AI provider.
Documents come in by upload or from a folder on the server.
Built with Python, FastAPI, Next.js and SQLite full-text search.
Status
Plancite is in pilot and not generally available. A full demonstration on synthetic data is available on request.
Not done yet
Reading real-world benefit documents. Extraction handles Plancite's own document layout today.
Connections to more document sources beyond upload and a server folder.
Real member eligibility and accumulators.
Testing with large document libraries and many users at once.
Regulatory and utilization-management review.
Interested in Plancite?
We'll walk you through Plancite on synthetic data. Tell us how your team answers these questions today, and we'll reply within a business day.