Two-minute savings per call is just the beginning: Spark transforms workflows with Microsoft Copilot

Anyone who’s spent time in traffic jams will appreciate how one small incident can end up creating much bigger flow-on impacts affecting the whole network. Spark is one of New Zealand’s largest businesses, having grown from a telecommunications provider to a full-service technology solutions partner serving millions of individual and business customers. At such a large, complex business, resolving a single customer interaction is often a complex business in itself. Often, it involves multiple systems, teams and processes, with each step relying on people being able to access the right information and carry the enquiry forward quickly to avoid delays. 

In customer care, this can mean navigating several tools to answer a single query. In engineering and network operations, it often involves pulling together data from across complex environments, where even small gaps in data or delays between steps can slow delivery or increase the risk of providing incorrect information to partners and customers or worse, systems going down. 

At that scale, the challenge is not a single bottleneck, but the accumulation of smaller inefficiencies – repeated tasks, dealing with fragmented systems, and time spent moving between them. Together, these shape how quickly queries can be solved and how consistent the outcome is for customers. 

Having signed a major strategic partnership with Microsoft in 2025, which included the country’s largest ever Microsoft Azure cloud agreement and one of the largest Copilot deployments, Spark laid the foundations to completely transform and deliver better experiences to both its people and its customers. 

Identifying and solving workflow pain points 

For Jolie Hodson, CEO of Spark, and the wider team, the focus has been on understanding how work flows across the organisation and where it can be simplified. 

“We spent time looking at our end-to-end processes and identifying where work slows down or becomes more complex than it needs to be,” she says. “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.” 

In Spark’s contact centres, where thousands of interactions are handled each day, that complexity is particularly visible. Advisors are often required to move between systems to access information, validate details and complete follow-up tasks while managing the customer interaction in real time, and don’t always have immediate access to the technical information needed to resolve enquiries on their own. This leads to frequent reliance on the operations team, adding extra time to the process while also taking operational team members away from their day-to-day work. 

In response, Spark trained the Copilot tool with the technical and procedural knowledge call centre operators would routinely need, enabling more enquiries to be resolved at the first point of contact. 

Using Copilot as their foundation, the Spark team developed AI-powered assistants to cut out unnecessary steps. Rather than navigating multiple systems, they designed a solution that would present advisors with relevant information about the customer, products and services. The AI tool was also trained with information about Spark processes and procedures, so it could suggest next steps in real time. 

Throughout, Spark maintained a strong focus on governance, privacy and accountability, designing systems to make it clear who was responsible for making final decisions, with human oversight built in. While it was important to empower teams with more information, or automate certain processes, the defining philosophy was that team members, not AI, were in control.  

Success at scale 

AI-powered assistants are now supporting more than 350 frontline advisors, bringing together relevant information from across Spark’s systems and providing real-time guidance across billing, orders and roaming.  Advisors can now move more quickly from understanding a query to resolving it. 

The impact is both measurable and cumulative, with around 1.5 to 2 minutes saved per call through automated summarisation. AI assistants are now handling approximately 20,000 internal queries each month, reducing the volume of questions flowing back to operational teams by around 60 per cent.  

Hodson says these improvements matter because of how they scale. 

“When applied consistently, they reduce rework across the system and improve both speed and consistency,” she says. “It also allows our teams to focus more on resolving customer needs, rather than navigating the process itself.” 

The same approach is now being applied across other parts of the business. In business customer product set-up, automation has reduced the time required to complete processes by around 60 per cent.  

In software development, Copilot is supporting tasks such as drafting test cases, assisting with migrations and validating changes before they are merged – reducing time spent on repetitive steps while maintaining quality controls, and enabling faster release cycles. 

In network operations, AI is used to detect potential faults earlier and recommend steps to remediate any issues, supporting earlier intervention. That’s a massive potential boost not only to Spark’s productivity, but to that of its customers.  

Becoming a Frontier Firm of human and AI agents 

This is only the start of the journey. Spark is now exploring how AI can adjust energy usage across sites in response to real-time demand, saving costs as well as the planet. 

It’s also investigating how Copilot and other Microsoft Azure AI solutions can be applied across broader workflows, letting systems carry out defined parts of a process before handing over to human team members for approval. 

This marks a shift from AI helping people arrive at an answer, to AI progressing the work itself – planning steps, executing defined tasks, and handing over where human judgement is required.  

“We’re moving from AI that helps people get to an answer, to AI that can help complete parts of the work itself,” Hodson says. “It changes not just how quickly things can be done, but how consistently they are delivered.”  

Jane Livesey, President of Microsoft Australia and New Zealand, says this approach to AI deployment, which moves beyond simply doing the same tasks faster to transforming whole ways of working, is what enables organisations to scale AI more broadly. 

“We often see organisations get early wins from AI, surfacing information faster or helping people move through individual tasks more quickly. But on its own, that doesn’t always change how work runs day to day,” she says. 

“The real step-change comes when leaders look end to end: how work moves across the organisation, how systems connect, where decisions sit, and what it takes for teams to coordinate. That’s where AI starts to scale, because you’re not just improving one step, you’re reducing effort across the whole process. Spark is a standout example of doing this boldly and responsibly, and showing what meaningful ROI from AI can look like for customers and employees. I’m excited to see what they unlock next.”