The ambition is right and the technology is ready. What is not ready, in most cases, is the estate data underneath it.
The commitment is now national policy. The NHS intends to become the most artificial intelligence-enabled care system in the world. For estates and facilities the opportunity is not hypothetical:
- Predictive maintenance
- Compliance monitoring that flags risk before an inspection does
- Energy optimisation that responds to how buildings are actually used
All of it is technically available today.
What is less widely discussed is what the first year of that work looks like in practice. For most Trusts it will not look like AI at all. It will look like reconciling spreadsheets.
What these models need from estate data
A predictive maintenance model needs asset histories with continuity, which means more than a record of the last eighteen months. It needs enough years of failures, repairs and replacements for a pattern to be visible at all. Compliance monitoring needs inspection records that agree with each other, so that an asset carries the same identity in the maintenance system as it does in the compliance register. And anything touching space needs a floor area that means the same thing everywhere it appears. The same room must be measured on the same basis, with the same reference in the CAD layer, the estates return and the cost allocation.
Very little estate data was collected with any of this in mind. That is not a comment on how estates teams have worked. It is what happens when systems arrive one at a time across a decade, each procured to answer a specific question, each entirely coherent inside its own boundary. The incoherence is not within those systems. It sits between them, and nobody was ever asked to own it.
There is something else worth naming. Most estate data was gathered to satisfy a return or an inspection rather than to be analysed, and data shaped by a submission deadline behaves differently from data shaped for continuity.
The work itself
Getting from one to the other is fairly unglamorous, and it follows a sequence.
It starts with finding out what exists. Which datasets are held, in which systems, maintained by whom, and last verified when. This usually surfaces more sources than anyone expected, including the spreadsheets that quietly became authoritative because they were easier to update than the system they were meant to feed.
Then the harder decision. For each dataset, which source is the one that counts. That sounds administrative until two systems disagree and someone has to say which is right. Asset identifiers, floor areas, cost codes and condition scores are where this bites, and condition scoring is often the worst of them, because surveys carried out years apart by different surveyors against different scales rarely produce numbers that can be sensibly compared.
After that comes ownership. Who is accountable for each dataset staying accurate, and how disputes get settled when they arise. This is the part most likely to be skipped. However, it is the part that determines whether any of the cleansing still holds twelve months later.
Very little of that is technical. It is governance, and governance is not something a procurement exercise delivers.
Where we fit
This is the work we do with NHS Trusts, and it is the reason we built OPAL rather than buying a platform off the shelf. Most consultancies either advise or deliver but we do both. The advice is shaped by having done the reconciliations ourselves rather than recommending that somebody else does them.
OPAL’s Operational Performance module is where the cleaned baseline lands, holding space, condition, carbon and cost in one place so the agreed version is the version everyone works from. Asset Management keeps the registers consistent as the estate changes, which is what stops the reconciliation becoming an exercise you repeat every three years. The platform matters here, but it is not where this starts.
Do not wait for everything to be right
None of this argues for pausing until the whole estate is immaculate. That would take longer than the AI opportunity will wait.
The sensible approach is to pick the dataset that the largest number of decisions depend on and get that one genuinely right before moving on. For most Trusts that is space, followed by assets. Fixing space records properly improves utilisation analysis, cost allocation, capital planning, statutory returns and energy reporting at the same time, because all of them rest on the same foundation. Getting one dataset to that standard is more use than improving five of them a little.
The Trusts that see real value from AI in estates over the next few years will not necessarily be the ones that bought earliest. They will be the ones whose asset registers were consistent, whose maintenance histories ran deep enough to learn from, and whose space data everyone had already stopped arguing about.
We are happy to look at what your estate data would support today, and what it would take to close the gap before an AI investment is committed.
Contact us for more information
