When AI Meets the Ecosystem

In the last edition, we looked at where AI delivers return on investment inside a lender’s existing estate. Document classification. Income verification. Case summarisation. Practical, well-sequenced work that builds AI maturity and funds what comes next.

But the internal horizon is only half the picture.

The external horizon now has a delivery timeline. In the first four months of 2026, five separate programmes have moved from policy intent to active build. CFIT’s Open Property Coalition, convened by government, is transitioning from blueprint to delivery this month, with a smart data scheme roadmap targeting a Digital Property ID that would bring together verified title, valuation, compliance, and chain-progress data into a single reusable record. The FCA published its Open Finance roadmap in April, naming mortgages as a priority use case, with a PolicySprint on mortgage data sharing underway in Q2 and a discussion paper on the first formal scheme due in Q4. The FCA’s AI Live Testing sandbox has moved into its second cohort, with Coadjute now testing AI-powered AML compliance specifically for the property transaction, alongside Barclays, Experian, and Scottish Widows. The Bank of England’s Synchronisation Lab has selected LMS to test atomic settlement of house purchases against HM Land Registry. And the Department for Business and Trade’s Smart Data 2035 strategy, published in March, names property as a priority sector and estimates £9.6 billion in annual GDP contribution from just four sector-wide smart data schemes by 2043.

None of these programmes is waiting for the others. All of them are converging on the same outcome: trusted, structured data flowing across the property transaction, between participants who have historically operated in isolation.

This matters for AI because it changes the operating environment entirely.

AI built for one organisation

Most of the AI being deployed inside lenders today operates on internal data, internal definitions, and internal workflows. That is the right place to start, and as we argued in the last edition, it delivers real return on investment when sequenced well.

But every one of those internal choices carries an assumption: that the data AI operates on will continue to look the way it looks today. A document classification model is trained on one lender’s application packs. An income verification model is calibrated against one set of definitions. A fall-through risk model is built on internal case-stage data that no external party uses.

These models work well inside a controlled environment. They are not designed to operate on data they have never seen, structured by organisations they have never worked with, under governance standards set by someone else.

That is precisely what the ecosystem now being built will require. Lenders who are investing in AI without considering the external horizon are building capability that may not survive contact with it. The design choices made internally today need to account for a data environment that is about to widen significantly.

Three things that change

Data standards will define what AI can do. The CFIT roadmap proposes eight smart data clusters covering the full home buying journey, from property identifiers and title ownership through to chain progress and post-completion records. These clusters will define how data is structured, labelled, and shared across the ecosystem. Lenders whose AI is aligned with those emerging structures will be able to consume and act on ecosystem data as it becomes available. Lenders whose AI is built on proprietary internal definitions will face integration costs and retraining cycles before they can participate.

This is a practical design question, not a theoretical one. The definitions your AI learns on today will determine how quickly it adapts when the data environment shifts.

Explainability becomes an external obligation. Inside a lender, the audience for an AI output is an internal team. The tolerance for opacity is a matter of internal policy. When AI outputs start crossing organisational boundaries, that changes. A risk assessment shared with a conveyancer, an affordability signal passed to a broker, or an AML flag raised in a multi-party transaction all require the receiving party to understand and trust what the AI has produced. Coadjute’s work inside the FCA sandbox is a live example of this being tested in practice, with AI-native compliance tools designed to meet the standards the FCA will expect when it takes over AML supervision of the property sector. The FCA has also confirmed it will publish a good and poor practice report on AI in financial services later this year. The direction is clear: AI that operates across boundaries will need to explain itself.

Governance has to travel with the decision. A mortgage transaction typically involves five or six participants. An AI decision made at one point in the chain, such as an automated valuation, a fraud signal, or an affordability pre-check, affects outcomes for participants downstream who had no role in producing it. The DRCF’s recent foresight paper on agentic AI raises exactly this question: as AI systems become more autonomous and operate across organisational boundaries, who is accountable for the outcome? Inside a single lender, accountability sits within the operating model. Across an ecosystem, it has to be designed in from the start.

What this means for lenders now

The temptation is to treat the external horizon as a future problem. The ecosystem is still being built. Smart data schemes are years from full operation. Open finance services are not expected to reach market until 2029 or 2030.

But the AI investment decisions being made now will determine whether lenders are ready when it arrives.

The practical question is whether the AI you are building internally is designed to compound into the ecosystem, or whether it will need to be rebuilt once the data environment changes. That is a design choice, not a technology choice, and it is one that Horizontal Digital Integration is specifically built to address. HDI connects the internal operating model to the external ecosystem, ensuring that the data, definitions, and workflows inside a lender are structured to interoperate with participants across the transaction. AI built on an HDI foundation is AI that travels. AI built on isolated internal definitions is AI that stays where it is.

Tools like Pathfinder by Novus help leadership teams model this. By simulating the impact of AI use cases on speed to offer, speed to completion, and fall-through, Pathfinder lets lenders see whether a specific investment creates value that compounds across the journey or value that is locked inside a single process step.

The five programmes listed at the start of this article are building the infrastructure for AI-first lending at ecosystem scale. The lenders who will move fastest when that infrastructure matures are the ones making design choices today with the external horizon in mind. Not because the ecosystem demands it yet, but because rebuilding later costs more than building well now.

In the next edition, we bring both horizons together and look at what parallel execution means as a leadership discipline for AI investment.