Spark, New Zealand’s largest telecommunications and digital services provider, has over a decade of experience embedding artificial intelligence into its operations to drive productivity and revenue growth.

But the generative AI revolution, driven by powerful large language models (LLMs), has given the telco’s AI efforts fresh impetus. It has led to the development of SparkGPT, a multipurpose AI agent based on ChatGPT that allows Spark’s 5200-strong workforce to use easy natural language queries to monitor sales trends, track the status of contracts, and much more, democratising data access and decision-making.

Modest beginnings, big impact

The real power of AI was demonstrated over a decade ago at Spark, when the data and automation team was tasked with tackling the problem of “churn” in pre-paid mobile customers, 30-35% of whom were leaving the telco after a short period.

“Using several sources of data relating to a customer’s usage patterns, we could employ AI to predict when a customer was likely to leave,” says Anshuman Banerjee, General Manager, AI and Data at Spark and an 11-year veteran of the company.

“That allowed us to have a chat with the customer or provide an offer to retain their business.”

It took nine months to create the AI model underpinning the tool, and three more to convince the business to use it. However, the payoff was significant, resulting in a 20-30% reduction in churn.

A more recent AI-driven transformation has taken place in Spark’s contact centres. Traditionally, agents spent two minutes after each call manually documenting the discussion.

“Now, within three to four seconds of the call being completed, AI generates an automated summary based on the call transcript,” says Banerjee.

“We’ve saved around two minutes per call, and our Net Promoter Score, which measures customer loyalty and satisfaction, has improved because our people spend more time on calls, having better, higher-quality discussions.”

AI was also drawn on to help staff at Spark’s network of 26 Business Hubs around the country better serve small and medium-sized customers. While Spark account managers would typically contact customers regularly to check if their needs had changed, a predictive AI tool drawing on geospatial data allowed Spark to establish which customers were likely to be open to considering a new deal.

“It gives our Business Hub staff a view of acquisition, upsell, and cross-sell opportunities in their region,” Banerjee explains. “It has led to better customer service and contract re-signs.”

The productivity wins are real

These projects and others underway at Spark back up the findings of Spark-commissioned research by the New Zealand Institute of Economic Research (NZIER) published last year, which showed use of advanced digital technologies like AI could result in a 20% uplift in New Zealand’s economy and increase industry output by up to $26 billion over the next decade.

So, how can our businesses start reaping the productivity gains possible through embracing AI? It starts with people. Spark’s AI transformation has involved a major focus on lifting the AI skills and capabilities of its workforce.

Spark initially gave 700 of its staff access to Microsoft CoPilot and saw high engagement with the AI assistant of around 90%. After expanding availability to 2400 staff, engagement jumped further, to 92%. Banerjee attributes this to robust training and the use of AI ambassadors across Spark. 

“Almost every week, there is some training programme happening with partners like Microsoft, Snowflake or AWS,” he says. 

He emphasises the need to tailor training to specific roles: “If you’re talking to sales, it can’t just be ‘look what AI can do’, it has to be relevant to them.”

Resistance stemming from fear or stigma about AI is addressed candidly. 

“We had a colleague who used AI all the time and who said to us, ‘sometimes I feel like I’m cheating’. And we said, ‘No, keep going, we would love you to use AI like this.’” 

By framing AI as an enabler rather than a threat, Spark is fostering a positive culture of AI experimentation. It is among the first telcos globally to embed AI at scale across its operations, a position underpinned by a deliberate focus on responsible innovation, with strong data governance and ethical use evolving alongside new tools.

Towards agentic AI 

Spark is already employing agentic AI, which moves beyond information-serving bots to assistants that can undertake tasks on your behalf. A recent launch for an SME product condensed a process taking four weeks, requiring multiple handoffs and third-party involvement, into a single client meeting. 

Using a chat interface, call flow diagrams and contracts are generated in real time, enabling faster revenue recognition and improving customer perception of Spark as a digital leader.

Banerjee envisions AI’s role expanding end-to-end. “If a customer has a problem, can we take a lens back to our physical mobile and fixed-line networks? AI is beginning to enable the era of self-healing networks.” 

Summing up his advice for other businesses on their AI journey, Banerjee cautions against “big bang” AI transformations. Instead, he recommends starting with targeted use cases that have aligned leadership support, aiming to launch a real tool, not just a proof of concept, and measuring impact rigorously. 

“We have a dashboard showing the incremental value each AI project generates. It’s based on control groups, academic research, and user surveys,” he says. 

Basic AI – an absolute necessity

For example, time saved in contact centres and upsells from SME channels are reported regularly for everyone to see. “If you have good quality data, use it,” Banerjee stresses. Ethics and security are critical, but must evolve alongside innovation.

Finally, Banerjee highlights the human side: “Some people see AI as either a threat or cheating. But leadership and role models showing how AI helps in daily work can shift this mindset. AI isn’t the end of jobs, but it will create new, higher-quality jobs.”

“Basic AI is an absolute necessity now. If you’re not exposing proprietary knowledge and combining it with AI, you will fall behind,” Banerjee concludes.

Spark NZ’s AI journey is still just starting. But it illustrates how New Zealand businesses can embed intelligence throughout their operations, their culture, and their future ambitions.