Your best people should not be moving information around
Construction does not have a productivity problem so much as a repetitive-work problem. A large part of the week in most contracting businesses goes on experienced people acting as human interfaces between systems that will not talk to each other.
A quantity surveyor exports costs into Excel, reformats them and prepares another report. A site manager downloads photographs, renames them and uploads them somewhere else. A bid director searches old projects for a data sheet that is definitely somewhere. A managing director asks someone to pull together information the business already holds.
None of that builds the business. All of it is done by people who were hired to do something else.
What are we actually paying senior people to do?
Move information between systems, more often than anyone would admit in a board meeting.
The pattern is easy to miss because every instance is defensible. The report is needed. The photographs do have to be filed. Nobody is idle. That is why it survives: there is no single task big enough to challenge, only a hundred small ones adding up to a day a week.
The clearest case is the payment cycle, because it is mechanical and the consequence of getting it wrong is immediate. Subcontractor applications arrive, the due date runs, the payment notice falls due, and a pay less notice has to be served before the final date for payment. Miss it and the notified sum becomes payable in full, which is how smash and grab adjudications happen. That is a six figure exposure created by a calendar and an inbox, and there is no judgement anywhere in the calendar part.
Did we promote our best people into administration?
Partly, and by accident rather than design.
Construction promotes its strongest operators and then hands each of them a reporting obligation. Each is reasonable alone. Stacked over a career, they quietly change the job.
| Role | Promoted for | Where the week increasingly goes |
|---|---|---|
| Project manager | Running a job and holding a programme | Progress reports, chasing updates, RFI and TQ logs |
| Project quantity surveyor | Commercial judgement under pressure | CVR assembly, reconciliation, reformatting cost data |
| Preconstruction director | Estimating, design management and buying at bid stage | Tender documents and bid administration |
| Construction director | Fixing projects that are drifting | Email triage, board packs, governance returns |
| Managing director | Deciding where the business goes | Management information, and asking others to gather it |
Be careful with the top two rows, because the honest version is less dramatic than the pitch. A construction director does not sit updating spreadsheets; they have people for that. What actually consumes the week is meetings, travel between sites, email triage running to hundreds a day, and the two jobs in trouble.
Where the load has genuinely grown is governance. Monthly business reviews, gate reviews, risk register updates, framework KPI returns, ISO audit evidence, social value and carbon reporting, payment practices reporting twice a year, and now Building Safety Act competence records. Most of that did not exist when today's directors trained, and almost none of it needs their judgement. It needs their signature, which is a different thing.
The pain is also worst in the middle rather than at the top. The project QS, the document controller and the bid coordinator are drowning, and there are far more of them than there are directors.
Is this a British problem?
No, and the international evidence is older and blunter than the UK's.
The United States priced this in 2004. NIST's study of inadequate interoperability in the US capital facilities industry put the cost at $15.8bn a year, naming manual reentry of data, duplication of business functions and continued reliance on paper as the inefficiencies behind it. NIST called that a conservative figure. The scope covers commercial, institutional and industrial buildings, excluding housing and roads, and the number is in 2002 dollars and has never been updated. Take it as an order of magnitude rather than a current price.
US contractors have since reached the same conclusion themselves. The Associated General Contractors of America's 2026 hiring and business outlook reports 45% of firms deploying AI for office and administrative functions, against 23% for estimating. The largest use case is paperwork, by roughly double.
Australia has gone furthest in naming it. The Australian Constructors Association argued in Disrupt or Die that the biggest opportunity to improve industry productivity lies in how projects are procured, delivered and governed, which it describes as essentially white-collar activities, and notes that white-collar occupations account for the majority of the industry's skills shortage. Australia's Productivity Commission then put a number on the paperwork specifically: administrative and assessment costs make up around 33% to 36% of the total regulatory cost of building an apartment.
Worth noting what that does not prove. The Productivity Commission is explicit that weak construction productivity is not uniquely Australian. And in the US, Bureau of Labor Statistics economists have argued the opposite of the popular narrative, writing that the evidence from the individual industries they measured provides little support for the hypothesis that productivity growth has been abnormally slow in construction. They cover about 11% of construction hours, so it cuts both ways, but anyone citing US statistics to prove construction is uniquely bad should read them first.
The claim that survives all of it is narrower and more useful: three countries have independently identified the same white-collar information load, whatever the aggregate productivity statistics turn out to say.
Why is "it saved Bob six hours" the wrong way to value it?
Because it prices the hours at what they cost, and the interesting number is what they could earn.
Take a director working a fifty-five hour week with ten hours going into reports, chasing, packs and routine internal questions. The weak calculation multiplies ten hours by a salary, arrives at a few hundred pounds a week, and stops. That number is real and it is the least interesting thing about the change.
The real question is what those ten hours become. Walking a project that is drifting. Sitting with the commercial team before a CVR goes the wrong way. Reviewing a tender properly. Meeting an architect. Resolving a problem while it is still a problem and not yet a dispute.
Any one of those can win a substantial project or prevent a substantial loss, and none appears in a time-saved calculation. The arithmetic of hours returned is defensible and checkable, which is why it is the right place to start, but it undervalues senior time by ignoring what that time can produce.
The honest corollary, which vendors tend to skip: this will not reduce your overhead in year one. It is capacity, not cash. It lets a business run more work with the team it has, and if you want it to show up in the P and L, someone has to decide in advance what the recovered hours are for.
What should be automated, and what must never be?
Automate the work around the work. Leave the work itself alone.
- Do not automate the project director's judgement. Automate assembling the information the judgement needs.
- Do not automate the QS's commercial decision. Automate pulling together the evidence they need to challenge the number.
- Do not automate the client relationship. Automate the pipeline and follow-up administration, so somebody has time to go and see the client.
- Do not automate winning work. Automate finding previous experience, project data sheets, CVs and tender evidence, so the bid team can spend its time on why this client should appoint this contractor.
One caution on the boundary, because it is finer than it looks. A CVR is not cleanly separable into gathering and thinking: deciding what to accrue for a subcontractor who has not applied, or what to carry for a variation instructed but not valued, is commercial judgement exercised at the moment of assembly. The same is true of a board pack, where deciding how the difficult job is worded is the political heart of the exercise. Retrieval is the recoverable part. Anyone claiming the whole task is recoverable has not prepared one.
There is also a control question that decides whether any of this is allowed. A CVR feeds reported results and gets audited. Automate any part of its assembly and your finance director and external auditors will want a control framework and a named reviewer, not a case study.
What if the real bottleneck is other people?
This is the strongest objection, and it is mostly right.
A four-hour board pack is rarely four hours of typing. It is forty minutes of assembly, two hours of waiting on four people who have not sent their numbers, and an hour of deciding how to present the job in trouble. Automation does not make a project QS submit on time, does not make a subcontractor send RAMS a week before they start, and does not make a designer answer an RFI inside the contractual period.
The evidence says the same. RICS ranks inconsistent approaches by supply chain partners fourth of nine blockers to digitalisation, behind cost and effort, and lists disciplinary silos too. Those are people problems. The Australian Constructors Association goes further and identifies the mechanism: most digital tools rely on open sharing of information, while the contracts in use encourage withholding it, because information is useful for preparing and defending claims.
So the honest claim is narrower than "automate the reports". Chasing is the automatable part. A reminder that goes out on day three and day seven without anyone remembering, an ageing log that shows what is outstanding and who has it, an escalation that happens on a schedule rather than when someone loses patience. That does not fix a contract that rewards silence. It does remove the excuse that nobody knew, and it means the person waiting is not also the person chasing.
How would you find out whether this is happening in your business?
By counting, before buying anything.
The instinct when margins tighten is to ask people to work harder. The better question is what the business is paying good people to do that a computer should already be doing. For every recurring task: who does it, how often, how long, where the information comes from, where it goes, and whether it needs judgement or is simply movement.
Then the question most reviews never reach: if those hours came back, what would we want that person doing instead? A saving with no destination is absorbed into more of the same work by Friday.
Two warnings before you start. It is not an afternoon's work if you intend to ask the people who actually do the tasks, and you should. And anyone who suspects an automation review is a headcount review will give you defensive numbers, so say what it is for in writing before the first conversation.
The method, the scoring and the things it will not tell you are set out in the construction automation audit, along with a starting list of candidate tasks so nobody faces a blank page. It is worth doing before any tool is chosen, and a baseline taken afterwards is not a baseline.
What most businesses find is not a scandal. It is a slow accumulation: a reporting line added here, a spreadsheet inherited there, a pack somebody once asked for that nobody has questioned since. None of it was a bad decision at the time. Together it is why the most expensive people in the business spend their evenings catching up on the work they were hired to do.
That accumulation is what AI Metric works on: drafting documents from a firm's own library rather than from nothing, holding the chase and the sequence so it happens every time, and training the people who have to operate it. Construction Metric does the site end of the same job, turning what a team already sends during the day into a structured daily record.