Healthcare AI pilot
ChartAuth
The patient. The coverage. The next care step.
Pilot: fictional records only, not a commercial product yet
Access is by invitation. The demo asks for a password I share with invited reviewers.
The problem
Before a patient’s next care step, someone has to confirm the coverage and whether the service needs an authorization. The answer is spread across the chart, the plan’s rules and a payer portal, and a miss shows up later as a delay for the patient or a denied claim.
What the pilot does
ChartAuth is a working pilot of eligibility and authorization readiness. For each fictional visit it reads the patient, the coverage and the planned service, then gives a readiness assessment and a handoff: what is ready, what is missing and who should act next. It is built to decline a claim the record does not support, and automated tests check that behavior. The records and coverage rules are fictional, and the rules still need specialist validation.
What you will see in a 60-second walkthrough
The private preview screen, where the scope line reads “Eligibility & authorization readiness · Fictional records”.
One fictional visit, from the patient and coverage to the planned care step.
The readiness call, what is missing, and the handoff to whoever acts next, with the evidence behind each point.
How it is checked: tests that reject an unsupported claim, and where real rules would need specialist review.
Everything in the walkthrough is fictional. No real patient data is used.
Why I built it
At Optum Advisory Services I led clinical technology integrations for 20+ regional hospital systems and translated clinical requirements into technical workflows. ChartAuth takes that same handoff, between the chart, the coverage rules and the next care step, and asks how AI can make it ready without guessing.
Ask me to walk through it in an interview: the problem, the design decisions and what I would measure.
Design decisions I can walk through
Readiness, not a verdict
It says what is ready, what is missing and who acts next, instead of approving or denying anything.
Decline when the record is silent
A claim the record does not support is refused, and automated tests check that behavior.
Fictional records and rules first
Nothing real goes in until specialists validate the rules.
What a real pilot would need to measure
Not results. These are the questions a health system pilot would have to answer.
The share of visits whose readiness is complete before the visit.
The time from an order to a clear handoff.
How often staff confirm that a flagged gap was real.
What would have to be true before real use
Payer rules validated by specialists.
PHI handling and vendor terms agreed in writing.
A named human owner signs off each handoff.
Results compared with the current manual process on the same visits.
What it is not
Not a commercial product or a service offer. Not built on real patient data. Not clinical, legal or coverage advice.
Talk it through
For a walkthrough, ask for access. For Healthcare AI strategy work or a role, start with the options or the full profile.
contact@hurshdesai.com