Technology has long been pressed into service to solve the productivity riddle; but with the emergence of AI and robotics, infrastructure players are turning their attention to ensuring the payoffs are more than virtual.
The publication on 5 November 2026 of the British Standards Institution PAS 4010:2026 marks a significant moment for productivity in construction. Subtitled Improving productivity in infrastructure development and delivery, the standard is intended to tackle systemic efficiency losses across the lifecycle of projects.
The Productivity Management Strategy (PMS) at the heart of the standard requires organisations to implement an auditable, whole- lifecycle strategy, mapping frameworks for measuring and advancing performance. The standard also includes sections on accountability, overcoming individual project silos to encompass the construction ecosystem, from asset owners to contracting authorities, design teams and material suppliers.
PAS 4010:2026 sees AI and digital technology as productivity drivers and aims to guide projects via criteria including: data integrity and flow – which stipulates that organisations using AI for predictive design, digital twins or logistics must ensure structured data baselines exist to avoid feeding autonomous tools “garbage data”. “Human in the loop” thinking – with AI positioned as an informer of decisions to mitigate rework rather than an ultimate decision maker – is also key. The standard also stipulates balancing adoption with evaluating the factors aiding and hindering digital/AI integration across the project lifecycle, to ensure technology drives output rather than creating bureaucracy.
Production thinking
In the same vein, Mace Construct’s Build Smart, Build Better report published in June argues that robotics, AI and automation only deliver value when paired with earlier project decisions, standardisation, repeatable design and production-led delivery models.
The report includes earlier initiatives such as Project Lift – Mace Construct’s internal digital transformation and productivity programme, launched in 2025 to rethink how construction projects are set up, managed and reported.
Mace Construct director of technology Paul Connolly – who contributed to Build Smart, Build Better – says the report’s key message is that most productivity and innovation potential, including digital maturity and modern methods of construction, should be determined during early planning. One aim is to challenge the industry’s tendency to view automation as a standalone solution, instead making the case for a construction-to-production mindset, he says.
“We’ve spent many years looking at the dysfunctional, disruptive part of why we can’t improve,” he notes, adding that over-reliance on technology could be one part of that dysfunction. He cites BIM – building information modelling – as a classic example.
“When BIM emerged in the sector, the thinking was – if we model everything, we’ll get better. Well, that was what, 15 even 20 years ago? – but projects are getting longer.”
He believes part of the answer lies in bringing production thinking into delivery.
“Whether you design in a model, whether you design on paper, whatever project it is – a road or building, it comes back down to the starting point, standardisation. So, if you have data sets that you can use multiple times across multiple areas – a number of inputs that from the start set the culture and the behaviour – that underpins effective tools, whether it’s BIM, robots, AI or modularisation.”
Connolly stresses that the appropriate use of technology can be invaluable here.
“For example, you could pump into AI ‘give me a list of higher risk activities in construction’, whether that be road, whether that be rail, and you can say – those higher-risk activities in the current design, are they fewer or more with this way of building? You might then get the AI to recommend actions to influence a better outcome in terms of how you can reduce that risk.”
Physical AI has also been tested by Mace Group, including the trial of Boston Dynamics’ Spot robot dog in December 2022 at Heathrow Airport, to perform autonomous 3D laser scanning during the refurbishment of a cargo and luggage tunnel.
But Connolly argues that ultimately, systemic change is needed to ensure projects are set up to make the most of new technologies.
“Without talking about specific projects, I think that transactional baton-change between design, construct and commission handover has probably been one of the elements that affect productivity. I do think that is probably one of the areas that can be changed, so that your design will support the effectiveness of using automation, AI or robotics.”
Supply chain engagement
This mindset is reflected in the approach adopted by Costain, which regularly engages with tech-led Small and Medium-sized Enterprises (SMEs) or start-ups to advance innovation. In April, Costain engaged Hyperion Robotics to deliver 90 robotically manufactured, 3D printed low-carbon concrete sleepers for the flagship East Coast Cluster project on Teesside, in partnership with A E Yates.
The project requires the design and construction of multiple concrete pipe sleepers across the 1.3km pipeline, from the same material and of a similar size and shape, making it an opportunity for 3D printing at scale. It demonstrates benefits including the elimination of formwork and precise, repeatable production of the sleepers, requiring less soil excavation and reducing concrete and steel use by 40%. The results include a carbon saving of approximately 100t of CO2.
This approach draws on Costain head of innovation Rob Cairns’ experience during his previous role as senior innovation manager at HS2 Ltd, where he oversaw the use of AI to transform traditional infrastructure practices through production thinking – turning complex construction into an automated manufacturing process.
Initiatives included the HS2 Accelerator, to eliminate inefficiencies before boots hit the ground, and waste reduction through low-carbon trials, including the use of 3D concrete printing to cut tunnel concrete mass by up to 50%.
“That entire programme was all around de-risking technology,” Cairns notes, adding that to effectively de-risk a programme, the need “to really understand the need that the technology is serving, is super important”.
The Hyperion Robotics alliance notwithstanding, Cairns believes robotics are underutilised in construction due to a lag in industry adoption and mindset – and while autonomous technologies are being deployed on groundbreaking projects, the wider sector has yet to scale them.
“Look at farming and mining; robotics have been de-risked in those spaces for a number of years. So we have to figure out, how do we introduce that technology that doesn’t have a huge amount of people interacting with it into a site where you could have, tens of hundreds of people?”
Here and in terms of production thinking generally, deploying new ideas at scale benefits from the engagement of the whole supply chain, he says.
“The production thinking principles need to filter down to our supply chain through our contract onboarding, through our contract requirements, through our engagement, so that our suppliers understand what we’re aspiring to do to deliver a best-in-class service for clients.”
Through an alliance with Costain, Hyperion Robotics delivered 90 robotically manufactured, 3D printed low-carbon concrete sleepers for the East Coast Cluster project on Teesside
Powering up
AtkinsRéalis head of digital – nuclear Samuel Stephens agrees that new technologies should be embedded into real-world deployment, supported by strategic alliances, human-centric augmentation and stringent data foundations. Getting this right is crucial for the nuclear sector in his view.
“With the scale of the challenge to meet net zero, to decarbonise our society and then address all the additional demands from AI, we see a commitment to triple nuclear power by 2050, and the only way we will do that is through adopting new technologies and AI robotics.”
He says AI and nuclear power are inextricably linked by the huge power demand and requirement for low carbon base load energy to power emerging “AI factories” – powerful data centres that frequently scale past 100MW up to 1GW. “Forecasts say that there will be another 200GW of power required over the next decade and we see nuclear power as a key part of that.”
AtkinsRéalis’ collaboration with software and computer chip specialist Nvidia, announced in March, is a response to this, he says.
“One of the key technologies that we’re working with Nvidia on is the Omniverse solution, a digital twin platform that brings together 3D data from various information data sources into a common file format OpenUSD [Universal Scene Description]. That, with physics embedded into it, enables simulation of not only the construction but also the operation of facilities.”
This will help design, optimise and speed up the delivery of nuclear-powered AI factories.
But more broadly, nuclear schemes are suited to AI and robotics, given their technical complexity, high risk environments and need for rapid delivery through standardisation, Stephens says. He sees AI and robotics as emerging in the nuclear sector in three ways.
“Firstly, a lot of what we’ve seen to date and the productivity gains that we’re starting to see are through better information search and retrieval through generative AI. We have huge amounts of data in the construction and nuclear industries and using technology to get better insights from that is something that we’re grappling with.”
The second way is about the use of AI agents to support delivery. “That may be where we start to instruct and ask AI to perform specific activities under our direction and then report back. An area we’re looking at is licencing and permitting, where we have complex processes that we need to follow.
“We need to synthesise a lot of information and then report that back to developers and regulators to underpin the safety of the plant. If we can use those tools not only to synthesise information more quickly, but with a robust methodology and framework underneath and with humans to review and check the outputs, we see opportunities to accelerate construction.
“The third area that we’re seeing opportunities for, which is further down the line but increasingly relevant, is physical AI, where increasingly clients are interested in how they can use robotics and autonomous robots to carry out routine site inspections or walkdowns, both during construction but then also into operations.”
AtkinsRéalis is working with Laing O’Rourke (via its joint venture with Bouygues Travaux Publics) on the construction of the Hinkley Point C nuclear power station, creating factory concepts for delivering precast concrete to site, although that work primarily uses digital engineering, modelling and offsite manufacturing rather than physical robots.
But robotics are also increasingly set to play a role in industrialisation, supporting offsite assembly and factory-based assembly; AtkinsRéalis is developing autonomous nuclear robotics and remote operations for sites like Sellafield.
“We’ve worked closely to ensure that our designs are manufacturable in that type of environment rather than by humans on a construction site. That’s an area where we see potential for huge productivity gains.”
AtkinsRéalis is working with Laing O’Rourke (via its joint venture with Bouygues Travaux Publics) on the construction of the Hinkley Point C nuclear power station, creating factory concepts for delivering precast concrete to site
Scaling robotics
While Tier 1 organisations are rolling out innovative pilots across selected schemes, OLO Robotics founder and CEO Nick Thompson believes robotics remain underutilised in the UK.
“The UK is wildly behind the rest of the world [in robotics’ adoption],” he says. “Germany has got 20 times more robotic adoption than the UK. And China is so far ahead that we’re just miles away.”
OLO Robotics is positioned as a disruptor – supplying a browser-based software platform designed to support robot simulation, visualisation, control and programming without requiring specialist ROS (Robot Operating System) engineering skills. It uses AI to let users programme and control robots using plain English instructions via a built-in AI coding assistant.
Thompson believes the OLO software platform offers a cost effective and derisked pathway to adoption since the overall cost of investing in robot technology can be prohibitive. OLO Robotics enables users to purchase a price-accessible robot and deploy it quickly.
The OLO platform is agnostic, making it adaptable across a range of robotic technologies. “That means you can try things out, prototype and do things in simulation,” says Thompson.
He adds: “There’s a standout use case for robots, which is inspection robots, where a robot can be mobile, so it can roam around an area and capture data. Where they get interesting is when they’re autonomous because you can remotely control these robots or you can have them autonomous to a certain degree so you can give them high level instructions to go and do a job.”
OLO Robotics has attracted interest from “a large energy organisation” with facilities across the UK including electric substations, generation plants and gas works, he says. “These are dangerous areas, they’re also very large – and they lend themselves well to robotic inspection for faults or for routine maintenance and assisting people to transport things around autonomously without having to lay down tracks.”
To illustrate the potential productivity and cost gains offered by robotics software platforms, he gives another example – a soon-to-be announced collaboration with a well-known parcel delivery company.
“We went to the delivery hub and they explained their problem, which is they have these huge systems where the parcels are sorted. They come in one end, go on to a set of conveyor belts and they get zoomed around this distribution centre the size of 10 football pitches and full of little parcels flying around and falling onto the floor. That means they have to periodically shut the whole system down, pick them all up and put them back on again.
“Our solution will be to have a robot that goes around and cleans up, using the robot dog. That is a real use case and it will save them, over a course of a year, hundreds of thousands, if not millions, because of the immense disruption of having to shut things down and send in armies of people to clear these spaces up.
“It’s about the ability to be continuously operating. Every time they shut down, there’s a big cost to that and starting up again. So that’s what you’re able to mitigate and that is a huge amount of money saved. If you apply that to the water sector or energy sector or construction sites, you can see how it would work.
“In the construction industry, especially in infrastructure, there’s an unwavering demand for labour and for work and that’s not slowing down,” Thompson says. And with recent findings from digital construction platform PlanRadar showing that 70% of construction firms in the UK struggle to embed new technologies, he believes agnostic, user-friendly robotics software offers a de-risked route into adoption.
System reset
AI and robotics’ productivity promise is undoubtedly compelling. As Stephens comments: “AI is not going to solve everything, [but] where we will see real breakthroughs is where we can combine AI, robotics, automation and 3D design into standardised, repeatable, reusable digital assets and take an industrial manufacturing mindset to delivery.”
For Cairns, supply chain engagement will be key to ensuring that the whole infrastructure environment is ready to adapt.
“I think that’s where the Tier 1s bring that value to Tier 2 and 3, all the way down to the lower tiers of the supply chain – opening our doors to say, ‘If you’ve got something that could be transformational, let’s use our industry expertise, let’s use our sites and access to explore the opportunity. Let’s mature together so that we understand the risk, we understand the benefit’.”
While for now at least, human input isn’t going anywhere, in terms of the human-led systems deployed in construction, the need for change is urgent, according to Connolly.
He stresses that the infrastructure sector’s much talked-about collaborative models will need to become business as usual and much more widely, and authentically, embraced to take advantage of the step-change promised by these new technologies.
“It’s about bringing some of those key people on a lot earlier, through the design stages that will affect a faster outcome at the end, so that instead of having a linear approach, there’s far more integration earlier on,” he says.
“In terms of AI and related technologies, I think it needs to be a pure balance of how you use it and what you use it for. I don’t think it’s going to be the North Star for everything, but I do think it will play a part. I don’t think a single robot, or a single piece of software or platform will solve the productivity problem. These are enhancers, enablers – but we have to sort out the bits that disrupt progress at the start to enable their effectiveness.”
