The same engine.
A different lens.
The technical demo shows operators, reasoning chains, and Python dataclasses. This page shows the same engine running against the same data — without the code. Patricia Walsh, 74, Peterborough. Heating off six weeks. COPD. Her housing officer noticed before she had to ask.
The people
74 years old. Flat 3, 12 Longmoor Road, Peterborough PE3 8TN.
Widowed. Fixed income — state pension and PIP, £9,744/year.
COPD. Heating failed October 2025. Landlord repair delayed. Energy debt of £847.32 has accumulated. Prepayment meter credit exhausted.
EDF customer. Account 7731-4920-6.
Welfare and Tenancy Support Officer. Has known Patricia for three years.
He noticed the heating problem during a routine visit. He referred her before she had to ask.
Receives the AI-prepared summary and makes the human decision.
The engine tells her what it found. She decides what happens next.
The check
The EDF Customer Support Fund has five eligibility criteria and one exclusion check. The engine evaluates all six against Patricia's profile and returns a recommendation. Nothing is estimated. Every criterion is a direct comparison against a specific field in her record.
| Criterion | Patricia's record | Result |
|---|---|---|
| EDF customer | Energy supplier: EDF | — |
| Income below threshold | £9,744/year · scheme limit £14,000 | — |
| Has energy debt | £847.32 outstanding | — |
| Age 65 or over | Age 74 | — |
| Covered postcode area | Peterborough PE3 — within scheme area | — |
| Prior award in last 12 months | None on record | — |
The same evaluation runs in the technical demo where you can see every operator, the reasoning chain, and the raw Python output.
When a new scheme is activated
Eligibility doesn't wait for people to apply. When a scheme is activated, the engine can evaluate every known household against the new criteria immediately. The people who qualify are identified before anyone submits a form.
Below is the Ofgem Winter Warm Fund 2026 — a new scheme in this demo — dry-run against the three test applicants. In production, this would be the entire household dataset evaluated at the moment of activation.