As Women in BIM continues its AI Ambassadors interview series, Andrew Johnson shares his perspective on one of the biggest challenges facing the built environment, turning AI potential into practical, responsible adoption.
As Associate, Training and Development Lead at Okana, Andrew works with organisations to build the skills, confidence and governance needed to embed AI into everyday workflows.
In this interview, he explores the rapid pace of AI development, the reality behind industry hype, the capabilities professionals need to develop and how organisations can adopt AI responsibly. He also considers the balance between technological advancement and sustainable outcomes across the AECO sector.
AI capabilities are evolving at extraordinary speed globally. What are the most significant shifts you are seeing in how quickly AI is moving from experimental technology to embedded infrastructure?
The biggest shift is that AI has stopped being something you open and started being something you build on. Three years ago we talked about “trying ChatGPT”. Now models are built into the tools people already use, working in the background where most staff never see them directly. The pace is easy to underestimate. Google’s own figures show the energy per query fell 33 times in a single year while the answers improved. When capability climbs and cost falls that fast, adoption stops being a decision and becomes the default. For the built environment, the question is no longer whether to engage. It’s whether your people have the judgement to use what is already arriving inside their software.
Given this pace of change, where do you think the biggest misunderstanding lies among AECO professionals about what AI can already do versus what it will be capable of in the near future?
The biggest misunderstanding is about timing. Most people underestimate what today’s models already do well, drafting, summarising, checking, first pass analysis and overestimate the leap still to come, picturing an autonomy that isn’t close. So they wait for a future tool and underuse the capable one in front of them. A good example is the belief that the biggest, most powerful model is always the smartest choice. Often it isn’t: for everyday work a smaller, faster model matches what a heavier one produces, at a fraction of the cost and defaulting to maximum power can even make the answer worse. The skill that matters now is testing a model on your own real tasks, so you judge it on what it does, not on the headlines.
In the AECO sector, how effectively is AI currently being applied in practice and where do you still see the biggest gap between pilot projects, hype and meaningful day-to-day adoption?
Adoption is patchy. A lot of the sector is stuck in what I’d call permanent pilot mode: impressive demos, proofs of concept, a champion or two, but little that survives into everyday delivery. The tools work. The workflows around them often don’t, because nobody has redesigned the task, only bolted AI onto the old one. In some organisations the barrier isn’t capability at all; IT has switched AI off entirely over data and security concerns. The adoption that sticks shows up in small, everyday places: a bid team drafting faster, a dense specification turned into a plain checklist, a project manager turning a transcript into actions in minutes. The gap between hype and habit closes when organisations treat AI as a change to how work is done, with the training, permission and follow-up it needs.
You specialise in capability building and behavioural change. What are the most critical skills the AECO workforce needs to develop to keep pace with AI advancement and how should organisations approach this at scale?
This is my area, so I’ll be direct. The critical skills are not technical. They are judgement, clear articulation of intent and knowing when to trust output and when to challenge it. Anyone can be shown which button to press. Far fewer can frame a problem well, evaluate what comes back and own the result. A concrete example is knowing which model and how much “thinking effort” a task actually needs, and starting low, climbing only when the work demands it. At scale, the mistake organisations make is reaching for one-off events, a workshop or a webinar and calling it capability. Behaviour changes through deliberate practice on real work, with clear outcomes, support while they do the actual work and following up afterwards. Run another awareness session instead, and you get people who have heard of AI but still don’t use it well.
As AI becomes embedded in training, decision-making and delivery workflows, what risks do you see around over-reliance, deskilling or misplaced trust in automated systems?
Over-reliance is the obvious risk: people trusting polished, confident-sounding answers, even when they’re wrong. Deskilling is subtler and more worrying. If juniors let a model do the reasoning they should be learning, we erode the judgement the sector will depend on in 10 years. The answer isn’t to slow down. It’s to design how AI enters a workflow so it supports thinking and keeps a human accountable for decisions that carry risk. In our adoption framework, that assurance and governance step is a stage in its own right, not an afterthought. Used carelessly, it quietly erodes the expertise we most need to protect. Used well and with that governance in place, it raises the floor for everyone.
From a sustainability perspective, how should the AECO industry balance the environmental cost of AI infrastructure and compute with its potential to improve sustainability outcomes across design and delivery?
The debate here fixates on AI’s footprint and skips what it can save. Data centres use real energy and water, and the right response is to measure it and demand efficiency. But the built environment produces around 37% of global emissions, and very little of that is compute. It’s the concrete, steel and wasted material in what we design and build. Used on our models, AI can test low-carbon options early, cut rework, and improve how buildings run. The International Energy Agency suggests AI’s efficiency gains could offset much of the demand it creates, if adoption is widespread. So the balance isn’t AI versus sustainability. It’s whether people like you, working in BIM and digital construction, point it at the emissions that actually matter.
Hear more from Andrew at the Women in BIM Conference London on Thursday 15 October. Register today to join Andrew’s AI in Action Workshop.
