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 →
← Auditor console
ASM-0008LIMS · STAB-23-114-12MResearch & Development

Stability study 23-114, 12-month timepoint within limits

All attributes within limits; assay trend flat; no out-of-trend signals.

Open inspection evidence pack ↗

AI assessment

Simulated AI
B2
Risk level 2 · Formality B · RPN 9
Level 2 · AI review
Risk — severity × occurrence × detection
Severity3
Occurrence1
Detection3
RPN 9 → level 2
Formality — uncertainty + importance + complexity
Uncertainty1
Importance1
Complexity0
Formality B
AI confidence93%
The AI must be at least 75% confident to clear an item without a human.

AI’s reasoning (clear): Results sit within limits and show no adverse trend versus earlier timepoints.

Why no human was needed

Low risk, the AI was confident, and its grade met the formality ICH Q9 requires, so it was cleared at Level 2 and logged.

These reasons come from fixed rules, not from the AI. The AI can add reasons for a human to look but can never remove them.

Checked by the Assurance Layer

Checked, passedAll 8 Assurance checks on this proposal
  • ✓P1 · Scores are on the defined scalesAll six scores are whole numbers inside their scales.
  • ✓P2 · Confidence is a real percentageConfidence is 93%.
  • ✓P3 · Recommendation is one the platform allowsThe recommendation is “clear”.
  • ✓P4 · A reason is givenA reason is given.
  • ✓P5 · It rests on documents that are in force todayAll 1 named control(s) have a version in force.
  • ✓P6 · Supporting text was found2 supporting passage(s) found.
  • ✓P7 · Formality matches the riskGrade B2 is formal enough for the risk.
  • ✓P8 · The AI did not try to lower the barThe AI’s recommendation does not try to go below what the fixed rules require.

Rule set v1.0.0, judged on 2026-10-26. See the rules

Before the AI’s proposal is used, fixed rules check it. A failed blocking rule means the AI cannot clear the item by itself.

Grounded in the controlled documents

  • SOP-QC-009 v3.0 §4.3 Out-of-trend results
    A result that is within specification but departs from the established trend is investigated as out-of-trend before the next timepoint.
  • SOP-QC-009 v3.0 §4.2 Timepoint review
    Results at each stability timepoint are reviewed against specification and against the trend from earlier timepoints.

Only approved, effective versions are quoted, and only documents this record is mapped to. The assistant cites the same sections.

Controls and the regulations they satisfy

Company controlRegulation / guideline
SOP-QC-009 Stability ProgramICH Q1A(R2) · 21 CFR 211.166

History

Every step below is a record in the tamper-evident audit trail.

  1. Oct 26, 11:10 AMconnector.lims#36
    Record STAB-23-114-12M pulled from LIMS
  2. Oct 26, 11:10 AMai.simulated#37
    AI proposed a grade of B2 (RPN 9, 93% confident) — Level 2
  3. Oct 26, 11:10 AMassurance#38
    Assurance Layer (rules v1.0.0) checked the AI proposal: passed all 8 checks
  4. Oct 26, 11:10 AMsystem#39
    Cleared automatically at Level 2 — no human needed