Solving problems with AI and how to get started – with Dave Mackenzie

February 20 2025

Most organisations can see the potential that AI technology can offer but knowing where to start and how to unlock the most meaningful gains can be challenging. We talk to Dave Mackenzie, Aurecon’s Managing Principal for Digital, about the opportunities for New Zealand businesses with AI and how to get started.


It’s a challenging fiscal environment – why should clients invest in AI now?

“If you're not getting value from AI already, then you're missing out. There are so many ways to move the needle and increase efficiency quite simply with AI, and without huge cost.

“That said, it is a tough climate, so it has to be underpinned by value and impact. Clients need to be able to see what’s actionable right now. If they are unsure where to start, we tend to take them through Aurecon’s own AI journey, so they have something tangible to start from.

“One of the first tools we launched is Recall, which brings together Aurecon’s engineering expertise with AI technology to answer questions, and it’s unlocked productivity gains. We can see that, in the first 12 months, our people asked more than 301,000 questions – 26% of those were technical questions, 25% efficiency-based questions and around 23% skills development questions. So two-thirds of the questions our team are asking are business productivity based questions.

“We also have a construction request for information (RFI) tool, which has accelerated our response time by as much as 50%. On some of these major programmes, it’s totally transformational.”

Often clients know what AI is and understand the potential but have no idea where to start. Where is AI is helpful for technical businesses?

“There are three areas where we’ve seen AI offer huge value to clients: situations where they have big, complicated sets of data; they want to give more people in the organisation access to knowledge held by a small group of people; and when they’re working with asset management and operational data.”

Your clients deal with large volumes of traffic data, climate adaptation data and geotechnical data for example – is this the kind of thing AI can help with?

“That’s exactly right. Anywhere you have messy or unstructured data or information, that’s a great starting place for generative AI. Its superpower is processing unstructured data and turning it into something structured.

“We often say that data is the programming for AI, so having lots of good data is what drives great outcomes. Once that data is well structured, that can provide the foundation for more advanced things.

“Generative AI can analyse data like a person would. A model can take all of a client’s standards and guidelines and context, and become an expert in that organisation. Then we can use it to analyse large swaths of data, to generate insights and find new or unique connections.”

How can AI give more people access to specialist organisational expertise or understanding?

“Every organisation has someone who knows everything about your projects and knows all your standards and guidelines. They can answer every question, and they’re the kind of people that if they walk out the door, you lose a heap of knowledge, which can slow your organisation down.

“We bring together all your standards and guidelines, technical specifications – all the things your people need to do their job – and fine-tune a model specifically on those. If you think of ChatGPT as the bottom of the pyramid, this is going right to the top in terms of answer quality, and relevance to your organisation.

“So a transport agency, for example, might have a model that only knows about that entity. It will respond using the unique style and tone of voice of that organisation. Someone might ask it for the specific safety procedure for closing a road and filling a pothole for example, and it will be able to provide that, because your organisation is the only thing it knows about.

“This is a good place to start as the value is evident early on, it's easy to deploy and it’s relatively inexpensive.”

Can you share some examples of where you are currently working with clients using AI?

“We’re working on an exciting project with Transport for NSW (TfNSW), where we’ll be trialing ClaRFI, Aurecon’s tool that uses generative AI to interpret complex RFIs and collects relevant information from multiple sources within the database to provide guidance for responses. Given the productivity increases ClaRFI has created for Aurecon, we hope it will make a real difference for TfNSW.
“The goal of using ClaRFI is not just to automate the response process but also help the subject matter experts to find the information they need to complete the task quickly and accurately, so they can develop consistent and high-quality responses.”

Can you tell us about the asset management opportunity?

“Any kind of asset information is well suited to analysis using AI, and we’re starting to see clients looking for help towards programme and scheduling, as that connects to asset information as well. And certainly, different ways of prioritising asset decisions.

“The counterbalance to that is generative AI can create complex connections between data sets and information, which can really enhance decision making.”

How does the opportunity come about to work with a client on an AI solution?

“Technology must follow the problem, so we need to be developing tools in partnership with clients in response to real problems. Aurecon’s focus is on using AI to tackle complex problems that matter.
We’ll start by workshopping with a client a few potential use cases or problems their organisation is facing. Last time we did that, the client came up with 54!

“At that stage, the client will select a few of them – the high impact, low complexity ones – and we will quickly prototype something for them.

“With this technology, you can experiment and prototype easily, which means you’ll quickly know something is of value – typically within a day. On the other hand, if we start something and it suddenly becomes really complex, you’ll know that quickly too and can park it, without huge investment.

“We usually try to prototype a few ideas quickly with the client, and once they have something working, we can then expand that out with a pilot, and then onto a longer-term procurement.”

Can you share some other new or unexpected areas you’re seeing AI being used?

“One of the big areas coming up is multimodal AI. This is AI that has semantic understanding of images and diverse data sets, for example video, documents and text. AI can be applied quite simply to solve common problems – for example, getting manuals to people on site.

“It’s also helpful to get people aligned – to have a shared understanding. When you’re dealing with complex problems that require lots of people to collaborate quickly, you need systems and tools that facilitate that. With AI, you know you are talking about the same data, the same problems, the same language.”

Read more