IntakeAI / REVIEW WORKSPACEProject notes ↗

A SMALLER GAP BETWEEN A DOCUMENT AND A DECISION

Every field.
Back to its source.

Review an incident notice, inspect the evidence and prepare a checked summary.

Independent prototype
Synthetic examples only
No coverage or payment decisions
  1. 01 Read the notice
  2. 02 Check the extraction
  3. 03 Approve the summary
Recorded Python baseline

01 / Source notice

Select a field on the right to highlight its source. Document text is treated as data.

This example uses recorded output from the Python baseline. No live AI call is made.


02 / Extracted fields
Values remain editable. Corrections are recorded separately from extracted values.




03 / Review checklist
These checks follow illustrative demo rules, not an insurer's requirements.
    Awaiting human review.
    

    What this demo does and does not show
    The public preview replays real Python baseline outputs for four synthetic examples. It is not a live LLM. Free-form notices deliberately expose the limitations of label-based rules. The local Python server supports new text and an optional model adapter, which requires a separately configured API key and model.
    Source quotes prove literal occurrence, not semantic correctness. Review and export happen in your browser; there is no account, automatic claim decision or persistent server storage.
    Made by Abdirahman Jibril · Implementation & evaluation