You are looking at a read-only demo. Everything here is fictional and the AI is simulated. You can read every page and ask the assistant questions, but nothing can be changed.Sign in with an access code to try the full workflow →
How a record is handled

Ground it, check it, grade it, record it, release it.

Every record the platform handles takes the same five steps, in the same order. The counts below are live. Each step links to the page where you can see it working.

  1. 1

    Ground

    Find what the written procedures say

    A record arrives from a connected system. Before anything is judged, the platform looks up the controlled documents that govern that kind of event, and only the version in force today. If nothing supports an answer, it says so rather than guessing.

    Who decides: Fixed search over the document library. No AI judgement.

    14 of 14
    records have supporting passages from the documents
    Document library →
  2. 2

    Validate

    Check what the AI proposed

    The AI proposes a grade and a recommendation, or drafts an answer. Fixed written rules then check it: are the scores on the scales, does it rest on documents in force, is every quotation word for word. A failed blocking rule stops it.

    Who decides: The AI proposes. The Assurance rules decide.

    16
    AI outputs checked
    0
    stopped by a blocking rule
    Assurance Layer →
  3. 3

    Risk-tier

    Score the risk and decide who must look

    Fixed ICH Q9 rules compute the risk level (1 to 4) and the formality grade (A to D), then decide who has to look. High risk, low confidence, or a grade that is not formal enough always reaches a person.

    Who decides: Fixed rules. The AI can add reasons for a person to look, never remove one.

    4
    at risk level 1
    4
    at risk level 2
    3
    at risk level 3
    3
    at risk level 4
    ICH Q9 alignment →
  4. 4

    Record

    Write it down, together with the change

    Every step is saved in the audit trail in the same step as the change it describes, so nothing changes without a record. Each record is chained to the one before it, so any edit is detectable.

    Who decides: The database. It refuses edits and deletes.

    68
    audit records
    Intact
    chain, re-verified just now
    Audit trail →
  5. 5

    Release

    Clear it, or hold it for the right people

    Low-risk items the rules and the AI agree on are cleared and logged, and a person can still challenge them. Everything else waits for a person: the auditor decides, the QA head signs, and for level 4 an executive signs second.

    Who decides: People for anything that matters. Every decision carries a reason and a signature where required.

    7
    cleared by the AI under the rules
    7
    with a person or the team
    0
    waiting for a signature
    0
    decided and closed
    Approvals →

The one rule behind all five

The AI proposes. Fixed, written, tested rules decide. People sign. The AI is useful because it reads quickly and drafts well. It is never the last word, because its output is checked before it is used, its grade is recomputed by rules, and anything that matters reaches a person.

In this demonstration the AI is a placeholder with fixed outputs, and all data is fictional. The rules, checks, audit trail and signatures are real.