Four pieces on this site have each taken one part of the picture. Reality-Driven Intelligence set out how site reality is captured and verified, and why the order matters. The pretty model problem separated what a project intends from what a site is. A start-up owes you a number and What a day is worth show the one place the value of continuous evidence can be counted rather than argued.
This piece is about what holds them together - and why, of all the parties to a construction placement, Pikt built for the risk engineer.
Every fee on a project's insurance pays for paper, except one
Follow the money through the insurance of a large construction project. Premium pays for the risk transfer. The fees around it pay for work done on documents: brokerage for placing what has been presented, adjusting for settling what can be evidenced. None of that work requires anyone to stand on the site until something has gone wrong.
One fee is different. The risk engineering fee pays for someone to go and look before anything has gone wrong.
That is the whole distinction, and it is worth stating plainly because it is so easily overlooked. The other parties see the site at two moments: on paper at placement, and in person after a loss. Neither reflects the moment the risk is being run. Risk engineering is the function paid to look at the physical site while it is still only a risk.
The budget already exists - and it sets the frequency
The fee is agreed around placement, usually carried within the cost of the cover, paid out of the premium the insurers earn, and spent for the most part on visits and the reports that follow them. How many visits is a matter of arithmetic: divide the fee by the cost of a survey and that is the number the site gets. The budget sets frequency. Risk does not.
This is not a criticism of anyone. It is how a fixed fee works against a variable site. A project changes every week; the engineer arrives when the schedule and budget say the next visit is due, and between visits the site is unobserved by the function paid to observe it. What the engineer finds on the day is whatever has accumulated since the last visit.
Pikt sits inside that budget, not beside it. It is paid as a percentage of the risk engineering fee, for continuous coverage of the project - an existing budget line, not a new one. What the fee buys changes: it now buys the days between visits as well as the visits themselves. Every site in front of the engineer every day, changes flagged as they happen, and physical surveys directed at the projects where the evidence says they matter. The visits do not get fewer. They get aimed. More risk engineering from the same fee.
Why this function, and not the others
It would be possible to take continuous site evidence to any of the parties above. Pikt takes it to the risk engineer, for three reasons that turn out to be the same reason.
The first is that the risk engineer already reads the site. Once a project is running, the other parties' view of the risk they hold is largely the risk engineer's report. Put continuous evidence in front of the function that already looks, and it reaches the other parties to the placement by the route they already use.
The second is that, on the insurance side, the risk engineer holds intent and reality in the same hand while the project is live. The models and the programme say what was meant to be built by when; the site says what was. Reconciling the two is the survey. After a loss, others are paid to do the same reconciliation. Before one, that reconciliation sits with the risk engineer.
The third comes from the framework. Reality-Driven Intelligence describes a stack that runs from observation at the bottom to directing attention at the top, and its most useful claim is that the layers cannot be bought separately. Continuous capture with nobody reading it is a camera. Attention directed without evidence is a guess. For the insurance of a project, the risk engineer is where the stack closes: the function whose job is to decide where to look next, and then to go and look. Serve that function and the whole stack has a reader.
What reaches the other parties
This is the part that answers a fair question: if Pikt serves the risk engineer, what does anyone else get?
The answer is that the risk engineer's account is already one of the things the other parties rely on. Improve what stands behind it and the improvement travels.
For the insured, it means fewer reports to write and more to show. Where progress updates are already owed to insurers, they can be assembled from the site's own data rather than written by hand - a time-stamped account in place of a report someone had to compose. And if a loss occurs, the account of what the site was, day by day, exists already. It does not have to be reconstructed afterwards from fragments and recollection.
For the underwriter, once the project is running, the view of the risk they hold rests largely on the risk engineer's account. An account built on continuous evidence rather than periodic visits is a better-founded account. What any underwriter does with that is their decision, on their risk, and not ours to predict.
Where a loss occurs, the record of what the site was, day by day, already exists. Much of a delay claim turns on sequence, and a sequence that was observed does not have to be reconstructed. That is reasoning rather than a finding; What a day is worth keeps it labelled as such.
For the broker, the information is the same information, in a form that can be shown rather than described. What is presented, and how, is the broker's business and the market's, not ours.
This is not a separate product for each party. It is the risk engineer's account, better founded, reaching the people it already reaches.
Where the arithmetic fits
The framework says how the account is made. Risk engineering is who reads it and directs attention from it. The one thing neither of those settles is what it is worth, and we have been careful not to claim that it does.
For most of what is described above there is no number, and there will not be one - the severe losses are too rare to count and the softer benefits resist measurement. But for one line there is. Delay in start-up cover writes its sum insured, indemnity period, waiting period and rate into the policy before anything happens, which is why Reality Capture Value can show, on a project's own figures, how many days off the waiting period would offset the entire capture spend. Not a forecast - arithmetic, checkable by anyone with the policy schedule.
That number belongs in this picture for a specific reason. It is the insured's and the underwriter's arithmetic, not the risk engineer's - risk engineers do not set waiting periods. But the evidence such a recognition would rest on is the same continuous account of the site the risk engineer works from. The value is counted at placement. It is made on site.
Why this is the right way round
The market has tried more than once to get site data directly to the parties who price and settle construction risk. The clearest published example we have found is Shepherd, on US casualty, where the underwriter took the site data in directly and priced off it (Shepherd Savings, accessed September 2026; Engineering News-Record, 25 June 2026). Most of the London market is set up differently: its parties who price and settle read reports.
There is already a function set up to read a site. It has a budget, agreed on major placements. It has a reader, whose account the other parties already rely on. What it has not had is the site in front of it between visits.
That is what Pikt provides, and it is why the risk engineer is the customer.
See every site, every day. Survey the ones that need it.
Pikt - continuous site visibility for construction risk engineering. Pikt is an independent technology company. It is not an insurer, broker or MGA, and gives no insurance advice.