Quick Guide · Non-conformance and CAPA

Analyse monthly non-conformance trends with Copilot Chat

Isolated non-conformances closed across a busy month can obscure recurring mechanical or operational faults. Here is how to use Microsoft 365 Copilot Chat to cross-tabulate monthly defect logs by line and shift before technical verification.

14 Oct 2026 · 45 sec · Copilot (Copilot Chat, no licence needed)

The scenario

Site
Ambient bakery producing wrapped morning goods across two packing lines
Who
Quality Assurance Manager
Trigger
End-of-month review of the site non-conformance log showing 22 individual records signed off across the period
Why it matters
Individual incidents are dealt with promptly on shift, but systemic issues across shifts or lines can easily slip past unnoticed, risking repeat non-conformances at audit under clause 3.7.

BRCGS context

ClauseWhat the Standard expectsSensible practiceWhere technology helps
3.7
Corrective and preventive actions
Failures are investigated, root cause identified where needed, and actions implemented and verified.Sensible industry practice involves aggregating monthly corrective actions to check whether seemingly isolated minor faults indicate an unaddressed repeat cause.Copilot can rapidly group text descriptions and cross-tabulate counts by line and shift from an exported file; it cannot evaluate whether the recorded root cause was technically sound or whether actions were implemented effectively.

Clause intent in plain English, not the Standard's wording. Check requirements against your own licensed copy of BRCGS Food Safety Issue 9.

The approach, step by step

  1. 01

    Upload anonymised CSV log and request high-level categorization

    Review this non-conformance export. Group the 22 incidents into common technical themes and state the frequency of each theme.

    You check: Confirm that all 22 rows were included and that the theme boundaries reflect real production issues.

  2. 02

    Cross-tabulate repeat themes by operational variables

    For the sealing issues theme, break down occurrences by Line and Shift. Note any repeat patterns.

    You check: Cross-reference the raw export rows to confirm line and shift attribution matches source log entries.

  3. 03

    Evaluate stated root causes for repeat patterns

    Summarise the stated root causes for the Line 1 Night shift sealing failures. Did the investigations identify a common underlying factor?

    You check: Verify engineering job sheets and physical maintenance logs to validate whether an underlying mechanical fault was missed.

  4. 04

    Draft investigation focus points for the engineering and QA review

    Draft three factual review questions for engineering regarding Line 1 Night shift sealing performance based on these records.

    You check: Technical manager reviews questions for appropriateness before issuing to engineering.

Takeaways

  • Use Microsoft 365 Copilot Chat to cross-tabulate flat monthly incident logs by line and shift.
  • Identify recurring defect clusters that individual daily sign-offs can easily obscure.
  • Clause 3.7 requires technical verification; always substantiate flagged trends using engineering and operational records.

Limitations and risks

  • Copilot Chat does not connect directly to factory SCADA, ERP, or quality databases; data must be manually exported and uploaded.
  • The tool cannot verify whether an action was genuinely completed on the factory floor.
  • Free-text quality logs with inconsistent descriptions may lead Copilot to misclassify or split closely related defects.
  • Copilot does not possess context regarding recent engineering modifications or supplier material variations unless included in the file.
  • Accidental inclusion of sensitive site details, proprietary formulations, or identifiable operator names in uploaded files.
  • Accepting Copilot's thematic groupings without spot-checking against primary records, leading to incorrect CAPA prioritisation.
  • Treating AI pattern detection as a completed root cause analysis rather than the starting point for human investigation.

Sources

TFCD content is general guidance, not professional advice for your site. Always check requirements against your own licensed copy of the BRCGS Standard and your customers' codes of practice. Narrated with an AI voice clone of the presenter; Copilot screens are illustrative recreations with dummy data.