Construction
A one-day course for QSs, site managers, HSQE, engineers and planners. An AI reads the site file; your people decide what goes to the project director.
Status Built and in review. The first cohort has not run yet.
The day
Set on Ballymore Fields, a fictional 50-house scheme in Co. Mayo, in month three on site. In each exercise an AI tool reads a site file built like the real thing and puts its findings in front of the pair. The pair agrees or disagrees with each one, then makes the call.
01 ConditionsAI · 31 planning conditions
An AI reads the 31-condition planning register against the grant and the bill. Decide what goes into the cost plan and what the project director must close before the lender's drawdown.
What it teaches A works condition needs a cost, and "complied" means the council's written agreement is on file.
02 BillAI · 73 bill lines
An AI reads 73 bill lines against drawing revisions, returns, measures and utility offers. Price the changes that are real and note the ones that are not a cost.
What it teaches A revision is a change to price; a long lead time is a note for the project director, not a cost.
03 SafetyAI · 41-entry incident log
An AI rates a 41-entry incident log on potential, not counts. Pick the entries to escalate today and the ones that are trends for Thursday.
What it teaches Rate on potential, not outcome. Severity comes before counts.
04 SnagAI · 300 WhatsApp messages
An AI turns 300 site WhatsApp messages into a snag register. Issue it, but take out the safety and design items that are not a trade's to fix, and escalate them.
What it teaches A snag is one defect one trade can fix. Safety and design problems come out of the trade list.
In this course the AI tools are right about what they read. The skill is using them: checking each finding, knowing what the tool cannot see, and deciding who acts and by when. Marks go on the outcome, on how the pair used the AI, and on the reasons they gave.
Two AI personas sit beside the pairs all day. They answer from what is on the screen, never make the call, and say when a figure is not in the file.
How we measure the returnAI persona
Colm Kavanagh
Senior quantity surveyor. Runs the conditions and the bill. Brisk, exact, dry.
AI persona
Aisling Moran
HSQE manager. Runs the incident log and the snag register. Always asks what could have happened, not what did.
When trainees join, they tick what is on their job. The personas use the answers to bring each point back to the work they do on Monday.
The course is for quantity surveyors, site managers, HSQE managers, engineers and planners. We run it in a room with our iPads and our instructor, for one contractor's own people or for a training network.
Not yet. AI for Construction is built and in review, and its first cohort has not run, so there are no results to show. We will publish what the first cohort does.
Quantity surveyors, site managers, HSQE managers, engineers and planners at Irish contractors. Trainees tick their role, the work they do and the contracts they use when they join, and the AI personas tie each point back to that job.
It is stored in the EU, in Ireland. AI requests go through our server-side proxy to Anthropic, which processes them in the US under a data processing agreement and standard contractual clauses. Anthropic does not use API inputs or outputs for training by default.
An instructor runs it in a room, trainees work in pairs, and every exercise ends in a decision that is scored and debriefed. E-learning records that someone finished a module. Symerse records the calls they made.
No. No course can certify EU AI Act compliance, and we do not claim to. The course builds the practical AI literacy that Article 4 of the Act asks organisations to ensure in their staff.
A 30-minute call with the engineers who built it. No slides.