Data and governance
Trust is designed in, not added later.
A property record can only support decisions if every field carries its origin, its age and its uncertainty.
Commitments
Six trust commitments
Property identity
UPRN-first where appropriate.
Provider neutrality
Sources selected on value, rights, quality and economics.
Provenance
Show where data came from.
Freshness
Show when data was last verified or received.
Confidence
Make quality and uncertainty visible.
Human control
Decisions and actions must have transparent review and override paths.
Privacy and rights
Clear responsibility, clear controls
Data is selected, used and shared according to agreed rights, permissions, governance and operating controls. Property Spine supports clear responsibility for data quality, access and action.
Translate applicable mortgage regulation, Consumer Duty, financial-crime controls and data protection into operating requirements. Set evidence gates, accountable reviewers, escalation routes and AI oversight without transferring lending authority.
Agreed consultancy output. A responsibilities and controls plan with customer-outcome guardrails and an audit specification. Scope and responsibilities are agreed per engagement. Controls advice is not regulatory approval or a guarantee of compliance.
FAQ
Common questions
Is Property Spine a new data provider?+
No. It is a governed operating layer that can connect approved data sources into workflow and decisioning.
Does Property Spine replace existing platforms?+
No. It is designed to enhance existing workflow, data and operational capability.
Does it depend on one provider or portal?+
No. The architecture is provider-neutral and UPRN-first.
Can decisions be automated?+
Bounded evidence tasks can use automation under agreed controls. Portal AI is decision support, not lending approval. Authorised people review sources, apply lender policy and remain accountable for final decisions.
Responsible AI
Bounded tasks. Visible evidence. Human accountability.
Ground, challenge and override
AI can summarise supplied sources and flag risks, conflicts and missing checks. Reviewers inspect original evidence and lender policy, challenge suggestions and record the rationale for actions or overrides. A source reference does not establish that a model's interpretation is correct.
Govern before live use
Define permitted tasks, data minimisation, model evaluation, error handling and escalation. Test grounding, bias and harmful outputs. Agree accountable owners, access controls, monitoring and an audit record. AI must not issue an offer or determine lending suitability.
Consumer Duty
Speed must protect the customer and the lender's brand.
Design proportionate evidence requests, understandable explanations and appropriate support for customers who need it. Check whether faster processing changes outcomes for different case types or customer groups, not just the quickest successful applications.
Include repeat requests, avoidable waiting, communication clarity, support needs and process-related complaints in the baseline. Traceable records can help evidence customer outcomes and decision rationale. They do not guarantee Consumer Duty compliance or replace the lender's obligations.
Next step
Start with one lender pathway. Earn the right to scale.
Use buy-to-let remortgage property assessment as the example. Establish the baseline, engineer evidence and policy gates and test bounded AI under human control. Measure customer friction, decision quality and net cost as well as speed.