AI and predictive intelligence
Estimates that support professional judgement—never replace it
Predictive intelligence uses previous and current patterns to estimate where a problem may be more likely to emerge. It does not predict an individual's future with certainty.
Potential use cases
Where prediction could help
Persistent absence risk
Pupils whose attendance pattern may warrant review.
Attainment changes
Cohorts where progress appears to be shifting.
Workforce pressures
Recruitment, retention and cover pressure indicators.
Budget pressures
Where spend appears to be diverging from plan.
Contract overspend
Where commitments may exceed the agreed value.
SEND demand
Where provision demand appears to be growing.
School improvement trends
Where a set of indicators is moving together.
Limitations
What must be understood before any prediction is used
- Predictions depend on data quality
- Models can be wrong
- Bias must be assessed
- Individual context matters
- Predictions require professional review
- Appropriate legal and ethical assessments are required
- Access must be restricted
- Predictions should be monitored after deployment
- High-impact decisions must not be automated
Safeguarding boundary
Safeguarding decisions are never automated
Where lawful, appropriate and explicitly agreed, authorised information may help qualified safeguarding professionals identify patterns that warrant review.
- Restricted access
- Data minimisation
- Full audit history
- Professional oversight
- No automated safeguarding decisions
- No automated accusations
- No automatic family contact
- No unsupported risk labels
- No use outside the approved safeguarding purpose
EduGap is not marketed as predicting safeguarding incidents. Safeguarding case detail is never included in the public demonstration.
What does “Predictive analytics” mean?
Using previous and current patterns to estimate where a problem may be more likely to emerge.
What does “Machine learning” mean?
A way of finding patterns in previous information to make estimates about new information.
Request a MAT Data and AI Readiness Review
We will assess honestly whether prediction is appropriate for your trust yet—and what would need to be true first.