The record that has to prove itself
By Polly Barnfield, OBE, CEO of Maybe*
Maybe* builds AI that does real work inside businesses: the repetitive, reconciling, checking work that quietly consumes skilled people's time, done reliably and handed back for a person to approve.
That kind of work exists in every sector. But it takes a particular, high-stakes form in the built environment, and getting it right here has meant spending real time with the people who run these assets: the engineers, the asset managers and the operators who carry the responsibility when something is wrong, or cannot be proven right.
Spend enough time in that world and one pattern becomes impossible to unsee. The hardest problem in this sector is rarely a shortage of effort or competence. It is that the information an asset depends on is spread across systems, hand-offs and decades, and that someone, usually a skilled person who should be doing something else, has to reconcile it by hand every time a question is asked. That is the texture of the problem up close. And it is changing. This piece is about how, and about where the value moves as it does.
Why the sector settled for storing information
For decades the built environment stored its information and looked it up later. This was not negligence. It was arithmetic.
To verify a single drawing, against the standards in force the day it was issued, against the model, the specification and what was actually built, is close to a day of skilled work. A major programme holds millions of documents. An operational asset accumulates them for fifty years. Multiply the check by the volume, and the cost of verifying everything, continuously, was simply impossible. So the sector did the only rational thing. It stored the record and retrieved it on demand.
A stored record can tell you what was said, issued and approved, and by whom. It has never been able to tell you whether any of it was right, still true, or still safe. Those are questions about the world, not about your correspondence, and no filing system can answer them.
What is changing
The arithmetic is what is now changing. A layer that reads the record can also check it: continuously, on events rather than only when someone asks, and against the world outside the asset. The standard as it stood on the day. The ground. The condition of the structure. The state of the firms responsible for it.
Several forces are pushing the same way at once. The direction of regulatory reform is towards information captured once and reused many times, to a consistent standard, across an asset's whole life. In July 2026 the government published its response to the consultation on a single construction regulator, confirming it intends to create one and to move towards outcome-based regulation that treats buildings, products and professions as one connected system rather than separate silos. Implementation is years away, not imminent, but the direction is now settled. Alongside it, whole-life carbon is becoming a number that has to be evidenced, not estimated. Insurers and lenders are increasingly pricing risk on the quality of the record. And the simple productivity case is that the sector cannot keep reconciling information by hand forever. Every one of these rewards the same thing: a record in good enough order to be trusted and read.
Proof as a by-product, not a crisis
Whether the proof can be produced at all comes down to something basic. Can the people responsible see a single, current, shared record, and can anything check that record against the standards and the world around it.
Where information is fragmented, even a diligent, competent team cannot keep that proof intact. The gaps are not a failure of effort. They are a failure of arithmetic, the same arithmetic that made storage the only option in the first place. Where information is captured once and continuously checked, something better becomes possible: proof becomes a by-product of running the asset well, rather than a second job done in a panic when a regulator, an insurer, a buyer or a court comes asking.
The honest limits
It is worth being clear about what this does not do. Good information removes one obstacle. It does not remove all of them. Assets still fail, and projects still slip, for reasons that have nothing to do with data: competence, capacity, commercial pressure, funding. Better information is necessary, not sufficient, and anyone claiming otherwise is overselling.
And technology in this sector is frequently oversold. A system dropped on top of a team that was not helped to change how they work usually makes things worse, not better. The tools that succeed are the ones that fit how people already work, so the finished output simply appears where they already are, rather than asking them to go and operate another piece of software.
The shift worth naming
For years, the built environment argued about where its information should live. The real question now is what can check it against reality, and prove the answer.
Compliance and assurance should fall out of operating an asset properly, across its whole life, not be reconstructed from scratch every time someone asks. For the first time, information about a built asset can carry its own proof: every answer arriving with its source, its date, and the document it came from, and able to survive being tested years later, which is exactly when it will be.
