When I started thinking about writing a column on artificial intelligence, my first idea was to focus on the Bletchley Park summit held at the start of this month. Getting 25 countries and the EU to discuss AI safety felt like a useful milestone in the dizzying rise of technology like ChatGPT with its 200 million registered users.
And then I looked at the sector a little more closely. Within the space of a week, Microsoft had launched Copilot, its integrated AI assistant, Elon Musk had entered the fray with Grok, his “AI chatbot with a sense of humour”, and OpenAI had announced the opening of its GPT App Store. That’s a lot of progress.
This last one is probably the most significant step, given the way that the original AppStore turbocharged mobile technology development. This shift to a more varied AI landscape, with focused models trained on smaller sub- sets of data, could be a real challenge for the construction industry that sometimes struggles to organise and share its data to good effect.
The architectural, engineering and construction (AEC) sector has of course been using process automation and AI for years. Many design challenges cannot be solved without the power of genetic algorithms that learn as they solve.
Arcadis increasingly benefits from a growing cohort of citizen-developers using tools such as Python and Knime as well as large language models (LLMs) to exploit data more effectively. This enables us to create new insights and optimise decision-making for problems that our clients bring to us.
The industry’s wider adoption of AI is now more a question of timing than one of choice. We have no option other than to plan, prepare and act
However, the industry’s AI experts are at the tip of a much less capable iceberg. Non-specialist users like me have a different, arm’s- length relationship to AI. We happily accept recommendations from streaming sites, we occasionally use ChatGPT and we might be aware of the pace of change in the wider AI economy. But we use tools sparingly – partly because we have not yet developed the skills to prompt LLMs, but mainly because we still rely on pre-AI processes to form an independent point of view based on conventionally sourced evidence, knowledge and experience.
However, the industry’s wider adoption of AI is now more a question of timing than one of choice. We have no option other than to plan, prepare and act. For people at any stage of their career, that is quite a challenge.
The current speed of development means that, assuming that Microsoft and other providers get their pricing right, non-specialists like me will be using AI as a routine business support tool in the foreseeable future. Similarly, OpenAI’s App Store initiative raises the tantalising prospect of a new generation of niche, AI-driven apps focused on the needs of specific sectors such as commercial property.
What steps need to be taken to make sure that construction and property attracts AI investment and then has the capability to use it?
An underlying issue for all sectors is trust. Trust in data and AI applications was at the core of the Bletchley Park agenda. The existential risks around cyber security, misinformation and even biotechnology are so great that they have to be dealt with on the basis of a global compact.
Risk reports published before Bletchley highlighted that investment incentives are focused on scaling AI’s predictive power, rather than “conditioning” AI applications to perform in appropriate and socially acceptable ways. The joint declaration at the end of the summit should help to reset that balance of investment.
However, there are further trust issues associated with AI being used to support professional advice. Whose advice is it? Will it account for the specific circumstances of the project? Even basic considerations about how AI-supported advice can be assured have yet to be addressed when the advice comes out of a black box.
All of the construction, property and infrastructure professions will have a key role in addressing this challenge – particularly as the UK government has so far chosen to adopt a light- touch approach to AI regulation. Developing a flexible, proportional and risk-based framework that supports discretionary professional advice is an important pan-sector priority and I can see professional institutions such as the RICS having a key role in this.
A second and equally important consideration will be the AEC industry’s ability to scale its intelligence resource. LLMs are trained using the entire internet; generative design models can access tens of thousands of optimisation iterations from cloud-based servers; asset management systems can collect millions of data points.
But, in parts of the industry, data remains in silos, is transaction specific and is often protected by IP provisions. Access to large volumes of good quality, useable data is an age-old problem – particularly associated with commercial transactions and building performance. However, as data resources for solvable problems grow – contributing to better informed decision-making and assurance – the problems that have less data to support decision-making will become riskier and might attract less investment in AI as a result.
Our sector needs to keep on the data-rich side of the equation and, to facilitate this, an industry- wide approach to anonymous data sharing will probably be needed to train the AI. That is a big step, so we need to think about it sooner rather than later.
No industry will stand still and those that rely on analogue processes will be exposed to greater risk and will attract less investment and less talent
Trust and data are important issues but, without capable people to pave the way, construction’s route to an AI-powered future will be slow and uncertain. Elon Musk may believe in a world without work but, in the present, construction needs to create and retain a cohort of people with the capability and agency to develop the technology, processes, standards and culture to accelerate the adoption of AI, as well as the leaders to take these abilities to market.
Some but not all of these people will be technology experts. But to create this capability an industry response also needs speed and scale – speed to respond to pace of change and scaling because AI’s adoption will be needed in the industry’s SMEs as well as the sector giants who can afford to invest in training, policies and bespoke data sets.
I could not have written this column a year ago, yet such is the pace of change that the launch of an OpenAI app store is bringing the technology closer and closer to our work lives and everyday lives. Problems that can be solved with trainable data will be derisked and, as AI is adopted more widely in all walks of life, clients will increasingly expect AI-backed as well as professionally-based advice.
No industry will stand still and those that rely on analogue processes will be exposed to greater risk and will attract less investment and less talent. The only response to AI is to respond with scale and speed. Are construction and property professionals like us really ready for this challenge?
Simon Rawlinson is a partner at Arcadis