The Responsibility Gap

By David Black, Group Director for Growth and Development, Joule Group

Every few months an engineer asks me some version of the same question, usually one of the younger ones: will AI take my job? It is a fair worry, and the honest answer is more interesting than yes or no. AI is already changing how fire engineers work. What matters now is the choice in front of us about how we use it.

The fear of replacement misreads what our work actually is. Fire engineering has always sat at the meeting point of technical rigour and professional judgement. Calculations, modelling and codes give us a foundation, but a safe outcome depends on interpretation: on someone with experience reading the context, weighing the risk and deciding what the numbers really mean for a particular building and the people inside it.

AI is genuinely good at the first part. It can interrogate thousands of pages of codes and standards in seconds, accelerate modelling, surface patterns buried in large datasets and take the grind out of repetitive work.

At Joule Group, we have successfully used AI agents to carry out deep research to support a global benchmarking study conducted for a client in Saudi Arabia developing a Giga-Project. This research was highly specific with respect to global fire codes, legislation and performance-based design verification methods. What would have taken the team weeks was reduced to a matter of days. The agent did more than source the material accurately. Its review and summary of the comparative data gave our team a solid foundation for a high-quality deliverable, one the client credited with directly shaping their decision-making.

Those gains are real and worth having. But producing an output is not the same as standing behind it.

That distinction is the whole argument. In a safety-critical profession, someone has to be accountable. A model can generate fire and smoke scenarios in seconds; it still takes an engineer to question the assumptions behind them, check the inputs and decide whether the result describes a building anyone would actually choose to build. AI scales how much we can analyse, but it does not carry the responsibility for what we conclude.

We have been here before, in quieter ways. Computational fluid dynamics, advanced evacuation modelling and digital design tools each changed the daily rhythm of the profession. None of them made judgement less important; if anything, they raised the stakes on getting that judgement right. AI is the next step on the same road, and a larger one, because for the first time the tool can sit alongside the engineer as a working companion rather than a calculator. We already see it used for code interrogation, generating options, testing scenarios and drafting reports. Over the next few years, it will increasingly behave like a design co-pilot, suggesting routes through a problem, flagging risks earlier and widening the range of solutions an engineer can sensibly weigh.

The prize here is not speed. Fire engineering has never been a race, and shaving hours off a calculation was never the point. The real gain is consistency: a higher and steadier baseline of analysis across projects that grow larger and more interconnected every year.

The Gulf is unusually well placed to feel that benefit.

The region continues to deliver some of the most ambitious developments in the world, at a scale and pace you rarely see elsewhere, and increasingly through performance-based design rather than prescriptive code. You only have to open this magazine to see exciting updates on projects such as Etihad Rail, Al Maktoum Airport, Riyadh Metro and King Salman International Airport, as prime examples. This is exactly the kind of work where a tool that can test more scenarios and interrogate more data earns its place, provided it is used with discipline.

Which brings me to the part of this conversation that gets too little airtime. The risk I worry about is not that our profession will be too slow to adopt AI. It is that we will trust it too quickly.

Treat an AI output as an answer rather than a starting point and you do not get a safer building. You get false confidence, which is more dangerous than no confidence at all. The systems worth having are the ones that make an engineer more sceptical: tools that expose their own uncertainty, challenge an assumption and provoke a harder question.

Accountability is where this turns concrete. Clients, regulators and the public will keep expecting a named professional to stand behind a safety decision. AI cannot approve a design. It cannot defend a strategy in front of a civil defence authority or explain to an insurer why one approach was chosen over another. Those duties stay with qualified people, and they should.

Transparency follows directly. As AI works its way deeper into our process, an engineer must be able to explain how an output was reached, judge whether it holds up under scrutiny and set it out plainly for a client or a regulator. My own rule is simple: if a person cannot explain the reasoning behind an AI-assisted result, that result has no business shaping a life-safety decision.

There is an unglamorous side to manage too; bias, data quality, privacy and governance. AI is only ever as good as what it is built on, and poor data or weak oversight will quietly import risk into work that is meant to remove it. That is why I would argue firms should set their governance before they get excited about algorithms. Successful adoption has little to do with which licences you buy. It comes down to process, training, validation and oversight, and to embedding the tool within an engineering culture that already takes responsibility seriously.

This direction of travel is no longer hypothetical. When the NFPA launched NFPA LiNK in 2020, it brought its body of codes and standards into a single digital reference tool. At Intersec Dubai this January it went a step further, unveiling LiNK 3.0 with an AI assistant, CASI, that lets practitioners interrogate standards conversationally and pull cited answers in seconds. As codes become digital and machine-readable in this way, more of the job becomes interpreting intent, handling the exceptions the machine cannot, and turning dense analysis into a recommendation a client can act on and a regulator will accept.

The role of the engineer will shift with it. Technical competence stays fundamental, but the engineer of the next decade also has to be a capable user and governor of these tools, someone who understands where they are strong, where they fall over, and when an answer deserves to be challenged rather than accepted. The engineers who stand out will not be the ones who have mastered the most software. They will be the ones with the judgement to know when to put the tool down.

Underneath all of it, fire engineering remains a profession about people: how they behave when something goes wrong, how a building holds up under stress and how risk plays out in messy real-world conditions rather than on a clean model. Those questions are full of uncertainty, ethics and judgement, and no algorithm is going to answer them on its own.

So, my answer to the engineer worried about being replaced is this. AI will change how we work, and on balance it will make us better at it. What it will never change is who is accountable when a building has to keep people safe. The engineers and firms who thrive over the coming decade will be the ones who hold both of those truths at once: embracing the tool, while keeping a firm grip on the judgement, and the responsibility, that no machine can take from us.

Read more in David's latest Fire Middle East Magazine Ambassador article here: The Responsibility Gap