Authority Hub · Explainable AI
AI you can defend to a board.
Every recommendation is explainable. Every decision has a human.
Quick answer
Explainable Housing AI means every AI-assisted accommodation decision can be traced, audited and justified to residents, boards and regulators. Jigsaw Conferences uses AI to accelerate matching and reporting while keeping humans accountable for every placement decision — transparency first, automation second.
Housing decisions affect real people. When AI enters the process — matching residents to properties, prioritising placements, forecasting supply — the model must be explainable, auditable and human-supervised. Jigsaw's Explainable Housing AI position is the public statement of how we build, deploy and govern the AI in our platforms.
Written by Katie Richardson, Operations Director · Public Sector Procurement Specialist Last reviewed June 2026
What Explainable Housing AI means at Jigsaw
Explainable AI is not a marketing badge. It is an engineering discipline that requires every model output to be traceable to the inputs it saw, weighted, and the human review that signed it off. Jigsaw applies this discipline everywhere AI is involved in a housing or accommodation decision.
Our position is deliberately narrow: AI is used to accelerate professional judgement, not to replace it. The Jigsaw Suitability Framework — accessibility, family composition, safeguarding, cultural fit, community ties, health considerations — is embedded in the model. Every AI-assisted match presents both the score and the reasoning. No confirmation is fully automated: a person always signs off.
What matters here
Explainable
Every score is accompanied by the factors that produced it. No black-box outputs reach the buyer or the resident.
Bounded
The model only considers factors we have documented and defended. It does not consider proxy signals for protected characteristics.
Human-in-the-loop
No placement confirmation is fully automated. A named account manager signs every buyer-facing output.
Auditable
Every model call is logged. Inputs, weights, outputs, human review, decision — all recorded, retrievable and exportable.
Fairness-aware
Model outputs are regularly reviewed for disparate impact on protected characteristics under the Equality Act 2010.
Contestable
A resident, buyer or advocate can request the reasoning behind any AI-assisted decision. We produce it.
How AI is applied to a placement decision
- 1
Input capture
Human intake captures the buyer's requirements and the household's suitability profile. The model never invents inputs.
- 2
Model scoring
Candidate properties are scored against the suitability framework. The scoring rubric is public (see Resident Suitability hub).
- 3
Explainable output
The buyer sees the top-N shortlist plus the reasoning: which factors drove the score, which trade-offs the model made, what it excluded and why.
- 4
Human sign-off
A named account manager reviews the shortlist. Nothing goes to the buyer or the resident without a human signing off.
- 5
Audit record
Inputs, weights, outputs, review, decision and any appeal are logged. Records are exportable on request.
AI governance decisions housing teams face
Three moments where the explainability question becomes concrete — and what a defensible answer looks like.
If
Your organisation is considering AI in allocations or placements
Then
Require an explainability statement before adoption: what the model considers, what it excludes, where humans sign off. If a supplier cannot produce one, that is your answer.
If
A resident or advocate challenges an AI-assisted decision
Then
Produce the decision log: inputs, weights, outputs and the human review that signed it off. A challenge you can answer in a day is a system working as designed.
If
You are procuring software or services that embed AI
Then
Put AI due diligence questions in the tender: bias testing cadence, audit log access, human-in-the-loop guarantees, and DPA coverage for model inputs.
Due diligence checklistIf
The Ombudsman or an auditor asks how a placement was decided
Then
The file should show the suitability inputs, the scored options, the reasoning and the named person who approved it — whether AI was involved or not.
Suitability frameworkBlack-box AI vs explainable AI in housing placement
| Black-box AI | Explainable AI (Jigsaw position) | |
|---|---|---|
| Decision rationale | Score only — no visible reasoning | Score plus the factors that produced it |
| Ombudsman / audit evidence | Cannot reconstruct the decision | Full decision log: inputs, weights, review, sign-off |
| Resident trust | Decision feels arbitrary and uncontestable | Reasoning can be shared and challenged |
| Bias detection | Disparate impact invisible until harm occurs | Outputs reviewed for disparate impact under Equality Act 2010 |
| Procurement acceptance | Fails public-sector due diligence | Passes — explainability statement available |
| Human accountability | Diffused — "the system decided" | A named person signs every placement |
Procurement guidance
Buying AI-enabled housing services safely
AI is entering allocations, repairs triage and placement matching. The procurement question is not whether to allow it — it is what evidence to require.
- Require an explainability statement covering model inputs, exclusions and human sign-off points.
- Require audit log access: every model-assisted decision reconstructable on request.
- Ask for the supplier’s bias / disparate-impact review cadence and the last review date.
- Ensure the DPA covers model inputs — household data used for matching is personal data.
Downloads
Working documents for your team — everything in them is also published in full as HTML on this site.
Related services
Sectors we serve
Frequently asked questions
Does the model consider protected characteristics?
Can a buyer opt out of AI-assisted matching?
What happens if a resident disputes an AI-assisted decision?
Which AI providers do you use?
Related knowledge
Official legislation & guidance
Related articles & guides
Brief us once. We handle every detail.
Trusted by housing associations, insurers, local authorities, NHS organisations, pharmaceutical companies, corporate organisations and event professionals across the UK.