Healthcare

AI for chart review, prior authorization and intake, with every answer cited to the record.

Clinical and administrative staff spend hours reading charts to answer the same questions: what changed, what was tried, what the payer needs to see. DF1 builds assistants and workflows that answer from the record, cite the entry, and route anything consequential to a person.

Practice administrators, COOs, medical directors and revenue-cycle leaders at clinics, specialty practices, MSOs, home health agencies and revenue-cycle firms.

Five places to start

  • Pre-visit chart review

    A cited summary of what changed since the last visit: new results, medication changes, open problems. The clinician checks the citations instead of reading the chart.

  • Prior authorization packets

    The clinical evidence a payer asks for, pulled from the chart with entry references, drafted into the packet for review.

  • Referral intake

    Referral documents read, classified and turned into a structured record, with uncertain fields routed to staff.

  • Quality-measure abstraction

    Measure criteria checked against the chart, with the supporting entries listed for the abstractor.

  • Denial appeals

    The denial reason matched against the record, and an appeal draft with the cited evidence.

Document Intelligence, after a Discovery Sprint

Most healthcare engagements start with a Discovery Sprint on one workflow, using synthetic or de-identified data, and a Document Intelligence prototype over your chart format.

Discovery Sprint: two weeks, $4,900 fixed. Credited in full against a Build that starts within 60 days.

About Document Intelligence

Built for how healthcare works

  • PHI stays protected

    Demos run on synthetic records only. Discovery work uses de-identified samples or runs in your environment. Production work on PHI runs under an Enterprise agreement with a business associate agreement chain covering DF1, the model provider and the host.

  • Not clinical advice

    The systems summarize and retrieve from the record. Clinical judgment stays with clinicians, and the interface says so.

  • Every step logged

    Questions, retrieved entries, cited entries and reviewer decisions are kept in an audit log you control.

Running examples

Document Intelligence

Live

Chart Explorer

Questions over a patient chart, answered with numbered citations to chart entries, and "Not in this chart" when the record is silent.

Data
Synthetic: 25 Synthea patient records. No real patient data.
Limits
  • Not clinical advice and not a medical device.
  • Synthetic charts are cleaner than real ones; production work starts on de-identified samples.

Document Intelligence

Live

Blackletter

Answers questions over statutes and contracts with the source passage for every claim, and declines when the text does not support an answer.

Data
Public legal and regulatory text.
Limits
  • Answers only from the loaded documents; it does not know case law it was not given.
  • Every claim is provisional until a reviewer confirms it.

Data Platform

Live

Marketplace Data Platform

A migration verified by source-to-target parity tests, and a map of more than 35,000 seller-to-buyer trade routes.

Data
Public: a Brazilian e-commerce marketplace dataset.
Limits
  • A historical dataset; the platform shows the method, not current trade.
  • Analytics on a fixed snapshot, not a live feed.

Healthcare questions

Do you sign a business associate agreement?

Yes, for production work under the Enterprise plan, once DF1 holds the agreement chain with the model provider and the host. Until then, work uses synthetic or de-identified data.

Does it connect to our EHR?

It works from exports or interfaces your EHR provides. The Discovery Sprint confirms which one fits your system and your IT policy.

Is the demo real patient data?

No. Chart Explorer runs on Synthea synthetic records, generated for software testing. No real patient appears in any DF1 demo.

Is this a medical device?

No. The systems retrieve and summarize; they do not diagnose or recommend treatment, and they are built and labeled that way.

Tell us the workflow. We show you the system.

Describe the process and the data. A reply comes within one business day, from the person who would do the work.