Around AD 100, someone at the Roman fort of Vindolanda, in what is now northern England, needed to move stone. They wrote to someone who could decide how many wagons to send.
In the margin, a second hand asked for a reply saying what to do.
Now, I haven't looked too deeply into the dynamics between subcontractors and general contractors in AD 100, but the pattern is familiar. Someone on the job needed an answer. The information or authority to settle it sat with someone else. The question had to reach them, and the answer had to find its way back.
We've been chasing information on site for nearly 2,000 years.
The traditional way to find an answer
For much of construction history, the way to resolve a question on site has been to find the right person. Sometimes that person knows the answer, but often they have to ask someone else, check a plan, or look through a record before they can reply.
Today you probably just fire off a few phone calls and scroll through messages in the project WhatsApp group. You might run to your laptop and search through a drawing, or look through old emails. Either way, you're still chasing. The question can be answered, but someone's time has gone into finding, comparing, and relaying information that the project already holds. Time that could be spent doing better things.
From wooden tablets to Ctrl+F
Construction projects now have document systems, correspondence, programmes, site records and approval workflows. That has made information easier to store. But for all the software we have introduced, the biggest innovation in finding information on site sometimes feels like Ctrl+F. Someone still has to know which file to open, which term to search, and whether the result is the latest approved version.
You don't know if the answer sits in a drawing, an email, an RFI, or a meeting note. A search result may surface an older version or conflicting records. Someone still has to work out which source is current, what it says, and whether it is enough to act on. And that work usually lands on another person who is already busy doing their own job.
We went back to first principles (1,926 years back)
The first step should be as natural as the question itself. Walking onto the site, the stand-in supervisor asks aloud: “I'm covering for the supervisor today. What's happening on site, and what needs my attention before the crews start?”
From there, agents can do the chasing. One can check today's programme and planned work. Another can read yesterday's site diary and open actions. Another can look for recent instructions or unanswered RFIs. They bring the relevant records together and give the supervisor a short briefing with links to the sources. If the records conflict or an approval is missing, the system should say so and send the unresolved question to the person who can decide. It should never turn an absence of evidence into permission to proceed.
Yeah, voice is great for dictating a note or filling out a form. It's even better when it can chase up the information you need to do your job.
That writer at Vindolanda had to send a wooden tablet to another human and wait for a reply. We can now ask a question where the work is happening and have agents search the project's information for us. People will still make the decisions that belong to them. They should spend less of their day chasing the records needed to make those decisions.