AI that reads first, quotes its evidence and never decides
Before a reviewer opens an application, AI pre-review lists what may be missing or unclear, section by section — and every finding quotes the applicant’s own words. Reviewers adopt, edit or dismiss each one. The decision is always a person’s.
Three rules it cannot break
These are enforced by the software, not left to a setting.
It never decides
AI cannot set a section’s review status, record a verdict, recommend approval or release anything. Its findings live apart from the review until a reviewer adopts one.
It quotes the applicant
Every finding must quote the application. The quote is checked against what the applicant actually wrote, and a finding whose quote does not match is dropped.
A person is named
When a reviewer adopts a finding, the record shows the AI assisted and names the reviewer who accepted it.
How it fits your review
Every item here is a working part of the product today.
Checks before AI
Deterministic checks for missing elements run first, with no AI involved.
For reviewers
A pre-review panel per section: the finding, its criterion, the highlighted quote, and adopt, edit or dismiss.
For applicants
The same engine as a pre-submission check, so fewer applications come back for revisions. It is billed to the applicant, never the approver.
Bring your own key
Run every AI feature on your own model-provider key, under your own agreement with that provider.
Or a metered platform key
Use the platform key instead, with a cap and an estimate before each run.
Provenance
AI-assisted work is listed with the human reviewer who accepted it, in line with accreditor guidance that a named person reviews.
See a pre-review on a real application form.
Tell us who you are and what you run today. We will walk you through the parts that matter to you, using your own process as the example.