Overnight drone scans, automatic programme revisions to account for expected storms, and robots taking over from humans at the end of shift to work through the night – this is the vision of the construction site of the agentic AI future.
The vision is described in a white paper from Pi Labs, the construction and property technology venture capital firm that has secured £10m investment from Balfour Beatty.
In the white paper, Pi Labs states: “This is not a simple evolution of the construction site, but a step-change in how the construction industry operates. We are already seeing intelligent systems being applied to individual aspects of construction workflows in isolation. Agentic AI binds these together, enabling entire workflows to be automated.
“Rather than responding to past inputs from the user, agents can learn from historic data, monitor the results of their current actions, and plan ahead for future deviations.”
A typical day on site in the future
The present construction site
6am – the morning scramble
The site supervisor arrives and begins manually deciphering outdated schedules. Trades arrive, but due to a miscommunication, two different crews are scheduled for the same workspace, causing immediate bottlenecks.
9am – the data hunt
A project manager spends the morning frantically hunting through emails and paper blueprints for missing design specs.
11am – the unexpected disruption
A localised storm rolls in. The team scrambles to secure materials, and the day’s exterior work is abandoned. The schedule slips further, contributing to the 20-month average delay seen on a major project.
2pm – safety blind spots
A worker uses improper lifting techniques while carrying materials and water from the storm creates a slip hazard in a busy corridor. Both go completely unnoticed until a manual site walk is performed or an accident happens.
5pm – the hidden error
The crew heads home. During the day, a structural beam was installed three inches off its mark. This error won’t be caught for a month when the MEP contractors arrive to route ductwork, resulting in a costly change order and tearing down finished work, causing two extra weeks’ delay.
The day finishes at 5pm.
The future construction site
6am – the survey agent
Overnight, autonomous drones and ground rovers conducted a scan of the site. By morning, the survey agent has compared this physical data against the digital model and caught yesterday’s installation error today. The team corrects it immediately before other trades build on top of it.
9am – the site intelligent agent
An engineer queries the site coordinator agent in natural language about specific structural loads. The agent instantly retrieves the correct data from the digital twin, saving hours of manual searching and keeping the team moving.
11am – the scheduling agent
Predictive analytics flagged today’s localised storm 48 hours ago. The AI agent already dynamically adjusted the schedule, orchestrating just-in-time deliveries and re-routing trades to interior tasks, effortlessly avoiding downtime and reducing weather-related delays by 70%.
2pm – the safety agent
AI-enabled closed-circuit cameras and construction wearables act as an active safety net. The system instantly detects a puddle in a corridor and alerts maintenance, while simultaneously coaching a worker via their wearable on safe lifting techniques to prevent an ergonomic injury.
5pm – the automated construction agent
As human workers pack up, this agent orchestrates autonomous robots to continue the work. Computer vision validates the day’s work with 95% accuracy, and the site manager agent automatically generates compliance reports and field notes in minutes.
This cycle continues 24/7.
Jobs evolve with technology
Pi Labs describes how three existing site roles will evolve to match the needs of an agentic AI future and how a new role will be created:
- the project manager will become the AI programme director, shifting from chasing progress to overseeing AI systems;
- the site engineer will become the edge case specialist, focusing on the exceptions that AI can’t resolve (complex interfaces, ground conditions and design clashes, for example);
- the health and safety manager will become the AI compliance manager, as safety roles expand to oversee AI outputs, audit risk assessments and manage liability; and
- the AI prompt engineer is a new site-facing role testing and fine-tuning AI tools for project-specific workflows.
Pi Labs drills further into the impact of agentic AI, emphasising that technology will track data (including reporting and aggregation), while experienced site leaders will manage subcontractors, trust and conflict.
Tradespeople will need digital literacy to work alongside robotic and AI tools, while apprenticeships curricula will need to incorporate AI tools, data literacy and human-machine collaboration.
