DEMO AI SOLUTIONS AB
AI SYSTEM TESTING REPORT
Document ID: DEMOAI-RPT-TEST-001
Version: 1.2 | Approved by: Maria Lindberg, CTO | Date: 2026-04-01

=== FICTIONAL DOCUMENT — FOR DEMONSTRATION PURPOSES ONLY ===

1. TEST PLAN

Scope: End-to-end validation of Demo AI Classifier v2.1 against defined
metrics and probabilistic thresholds appropriate to the intended purpose
of automated defect detection.

Test schedule throughout development:
  - Unit testing:           2025-09 to 2025-10 (completed)
  - Integration testing:    2025-10 to 2025-11 (completed)
  - Validation testing:     2025-11 to 2025-12 (completed)
  - Pre-market testing:     2026-01 to 2026-02 (completed)
  - Post-deployment monitoring: 2026-02 onwards (ongoing)

2. TEST RESULTS

Overall status: PASS
Test date: 2026-02-15
Dataset: Production validation set — 12,847 images from 3 manufacturing lines

Metrics:
  Accuracy:            0.987
  Precision:           0.991
  Recall:              0.983
  F1 score:            0.987
  False negative rate: 0.017
  False positive rate: 0.009

Acceptance criteria:
  Minimum accuracy:            0.975 — MET
  Maximum false negative rate: 0.025 — MET

3. PRE-MARKET TESTING EVIDENCE

Full validation completed prior to placing on the market. Test results
documented and signed off by CTO on 2026-02-15. Testing performed against
prior defined metrics and probabilistic thresholds.

Testing methodology was designed specifically for visual quality inspection
use case. Test dataset composition mirrors production distribution of defect
types and component categories.

4. LIMITATIONS

- Testing performed under controlled factory lighting conditions only
- Novel defect types not present in training data may not be detected
- Performance on components from new suppliers not yet validated