Published on the 30/04/2026 | Written by Heather Wright

AI gains traction in big end of town

Don’t chase AI magic. Chase friction. That’s where the AI wins lie.

That’s according to Fonterra and Spark who are both selling the same idea in different uniforms: AI doesn’t need to transform everything overnight to make your P&L sing. At Spark, it’s as unglamourous – and as valuable – as taking two minutes out of a customer call and untangling the workflow knots that slow staff down. Over at Fonterra, AI is doing equally practical work: Monitoring butter packaging and pausing the line when faults show up, replacing spreadsheet-based production scheduling and pulling real-time IoT data from machinery across more than 100 plants into the cloud to support predictive maintenance.

“We’re using that database to pull research faster and to turn that into insights and innovation.”

Fonterra: From factory to office

Fonterra’s AI work spans both physical operations and corporate functions, with a focus on consistency, speed and decision support.

On the manufacturing side, the dairy co-operative, which processes around 22 billion letres of milk solids each season, is using AI at sites such as Clandeboye in South Canterbury to monitor butter packaging across multiple stages of production. When faults are detected, the system can pause the line, allowing staff to intervene early.

In production planning, AI has replaced spreadsheet-based scheduling, while real-time IoT data from machinery across more than 100 plants is streamed into Microsoft’s cloud to support predictive maintenance. The aim is to reduce downtime and disruption across a global manufacturing footprint.

At the same time Fonterra has embedded AI into everyday office work. The organisation was an early adopter of Microsoft 365 Copilot through the vendor’s early adoption programme, and identified hundreds of use cases across the business.

By February 2026, Fonterra says 35 percent of its global workforce was actively using AI tools, generating almost one million interactions in a single month across Copilot Chat, Copilot Studio agents, GitHub Copilot and the organisation’s own internal system, Co-op GPT.

The tools are being used to summarise meetings, capture actions, accelerate policy drafting and support internal decision-making.

Fonterra has also built a set of production AI agents using Copilot Studio in partnership with EY. These include an idea submission coach to improve the quality and consistency of investment proposals, an architecture assessment agent to support internal governance checks, and a technical accounting assessment agent designed to ensure financial assessments are audit-ready.

Skills, mindset and adoption at scale

Speaking at Microsoft’s AI Tour in Auckland last week, Fonterra CEO Miles Hurrell said the organisation’s AI work has been underpinned by a strong focus on upskilling its people and ensuring it has the right mindset around AI adoption. “There’s been a lot of work on how we get the right mindset, because you’ve got a range of views across an organisation,” he says.

Rather than positioning AI adoption as an IT-led initiative, Hurrell says Fonterra has taken a management-led approach. “We’ve gone with the approach of having it management-led as opposed to led from IT departments, and letting people explore and experiment in their own minds and share that knowledge,” he says.

One area where that approach is paying off is research and development. Hurrell says the company is using AI to improve outcomes at its Palmerston North research centre, which houses, decades of scientific work. With around 300 scientists, 120 PhDs and a century long legacy, the centre has a wealth of research documents. “We’re using that database to pull research faster and to turn that into insights and innovation,” he says.

Hurrell says the team is ‘hungry’ for the technology, but the organisation is focused on balancing experimentation with discipline. “It’s a matter of bringing the mindset and the skill set to ensure we don’t get too far ahead of ourselves, but experimenting and showing the benefit that we can create,” he says.

Spark: Fixing workflows first

At Spark the focus has been on understanding how work flows through the organisation and where it slows down or becomes more complex than it needs to be. Notes Spark CEO Jolie Hodson: “If you can remove some of that friction, you start to see a change not just in individual tasks, but in how the whole process comes together.”

The telecommunications and digital services provider signed a major strategic partnership with Microsoft in 2025, which included the country’s largest Microsoft Azure cloud agreement and one of its largest deployments of Microsoft Copilot.

The early impact has been felt in customer care. Speaking at the Microsoft AI Tour in Auckland, Hodson said Spark has more than 350 advisors now working with an AI system alongside them to help answer questions. Thousands of interactions are handled each day via the contact centres, and advisors don’t always have immediate access to the technical information needed to resolve enquiries on their own, requiring escalation to the operations team. Jolie says the system, a Copilot tool trained with technical and procedural knowledge call centre operators need, is deflecting 20,000 questions a month, or around 60 percent of issues, that would otherwise have gone to its back office.

The gains, however, are not limited to contact centres. Hodson says the organisation has deliberately spent time putting the right foundations in place before scaling its use of AI. “We spent quite a bit of time [on] the things you need to support the models – trust, privacy, the skills we need to build in our people,” she says. “With that you can then move at pace once the technology comes online.”

Spark is also using AI to accelerate its software delivery processes. “We are using it to test and change and still have human review inside that,” Hodson says, pointing to the use of AI to support development work without removing oversight.

She also pointed to broader ambitions for AI to lift productivity – a long term problem for Spark – and for the wider New Zealand Inc and Australia Inc. Hodson says AI is one way to lift productivity. “but standing back and waiting, hoping for something to change isn’t going to set us up to succeed.”

Moving from pilots to production

Across both organisations, the common theme is restraint paired with scale. Neither Spark nor Fonterra frame AI as a single transformational switch. Instead both talk about removing friction from existing processes, whether that friction sits in a customer interaction, a factory line, a governance workflow or research archive.

The signal is clear in how both organisations describe success: Measurable improvements in speed, consistency and throughput, grounded in skills, trust and governance rather than speculative pilots or one-off demonstrations.