AI is about better work, not fewer people, says Aurecon CFO Andrew Muller

Australia-based engineering design and advisory firm Aurecon, has put Microsoft Copilot into the hands of its 7,000 staff across Asia-Pacific following an 18-month trial involving around 500 users.

Aurecon’s Chief Information Officer, Andrew Muller, tells ADAPT that the Copilot rollout is about improving the quality of work the organisation can deliver.

Cutting costs and reducing headcount are not the goal.

“We’re a design firm, a people-based organisation, so the whole thing is about how we can use these tools to enhance what our people do. That’s really what we’re trying to do”, he says. 

Muller is a regular Copilot user, primarily using it to pull together complicated material, such as large internal reviews, and provide an initial view of the key issues.

“I use it a lot to give me a view of some pretty complicated reviews that we’re doing and ask, ‘Okay, give me the exec summary.’ But then, you’ve got to work with it, obviously.”

That caveat matters. Copilot produces a starting point, not a finished product, and the judgement still sits with the person using it, he says. 

In engineering design, quality is non-negotiable, and Aurecon runs a robust checking process before any work leaves the business.

For the past four years, the firm has also been using Nomic AI, a domain-specific platform for the engineering and built environment sectors. 

Nomic synthesises large volumes of technical materials in requests for information (RFIs) and runs an initial check of engineering work, identifying obvious issues before Aurecon’s own quality process takes over. 

“For us, quality is everything that we do. You want to make sure your product goes out the door and it’s right. [Tools] like Nomic can speed that process up. It can do a first pass of things, get rid of the obvious issues and really give you a first pass of a quality review”, he says. 

“We get massive requests for information, and it can synthesise those very well. But I think we are very careful; it’s not about getting a faster answer, it’s about getting a better one. Everyone quickly goes to, ‘Oh, we’re going to save all this money’. I’m not sure that’s the right approach for us. Enhancing what we do is what we talk about”, he says. 

Meanwhile, governance around these AI tools is simplified by Aurecon’s structure.

The organisation’s CIO reports to Muller, which means he represents technology at the group executive level. 

“I guess I can’t avoid thinking about the cost of this side of things. To be honest, because we’ve taken this approach of proving things up before we scale, we’ve been able to manage the costs from a technology point of view. 

“It [AI costs] will eventually replace a whole bunch of other costs…that [we] might end up bearing, whether through systems integration or software licensing. It will take over all of that”, he says.

The 15-minute test

Rather than build an elaborate business case, Muller set a deliberately low bar for both tools.

If each person saves 15 minutes a week and puts that time into client work, the investment pays for itself. 

“I can’t really work out the value use case with any certainty yet. I can have a spreadsheet that tells me the ROI and all these things, but we boil it down by saying, ‘If our people could save 15 minutes a week, it’s paid for itself.’ That’s good enough to start and then we see where we go from there.

“It’s all about keeping it simple. There’s no point trying to overthink the value because no-one has worked out what the value is yet”, he says. 

He calls the 15-minute saving a no-brainer and expects real returns to be far higher. 

“I’m expecting a lot more than that, to be brutally honest. We’ve got some tools that look at things and say, ‘The workflow that might have taken five hours for someone to do now takes minutes’”, he says. 

Aurecon has not experienced the AI token bill shock that some organisations are reporting. Muller says that costs have risen but remain manageable. 

“We’re pretty disciplined at managing our time, so to me it’s another cost we manage. At the moment, I haven’t seen cause to pause.” 

From hours to tokens

Copilot sits on a familiar subscription model, while Nomic works differently, charging customers through a combination of monthly subscription plans, a pooled AI usage balance, foundation model token costs and document indexing fees. 

For a firm that bills by the hour, the shift to consumption-based AI raises an awkward equation.

“If it costs me five hours to do this work, and now it’s going to cost me 15 tokens, what’s that trade-off mean?” he asks. 

Muller’s answer is that clients care about results, not inputs. 

“They want the value; they want the outcome. At the end of the day, it’s all going to be outcome based at some point in time”, he says. 

Muller remains optimistic. “I think it [AI] is worth it…the reality is that it’s going to disrupt a lot of business models and that’s just got to play its way out”, he says. 

What comes next

Aurecon has long favoured buying over building, with Workday as the core platform for its back office. 

“I like the idea of saying there’ll always be this core platform that will deliver the backbone of everything you do. You still need sources of record, and you still need the discipline that comes with that source of record”, says Muller. 

But he wonders whether AI tools will make it easier to build the smaller systems that sit at the edge.

He questions why the organisation would buy a big software license when it might not use all of the capabilities that the platform offers, pointing to CRM as an example. 

“I’d be more worried if I were one of these big incumbents at the moment”, he says. 

“We’ve got 7000 people. They are very smart. I’m sure they would build their own version of Excel if you let them, just so it could work a little bit more like what they want.” 

Staff are already creating their own tools. The challenge for Aurecon is spotting which experiments are worth scaling and then deciding whether to buy or build, he says. 

In finance, Muller expects agents to build on existing automation. 

“We’ve had the bots doing invoice processing for years. They’ll only get better now, so I can still see a world where we will have agents doing that sort of stuff.” 

He sees the biggest opportunity in removing the administrative layer, from timesheets to leave requests that require people to juggle diaries, managers and HR systems. 

“If we can make it easier for people to navigate the systems and find what they need, then absolutely there’s a massive play for that. The systems have still got to catch up, but I think there’s a big opportunity”, he says.