AI’s role in telecom sector growing, but customer satisfaction lagging

TORONTO – Nearly half of all telecom customers say they won’t use artificial intelligence for any of their phone and internet service needs, but others say they are open-minded as long as it involves a low-stakes task, like adjusting settings or setting reminders to pay their bills.

That’s the upshot from new global research by Ipsos Canada, shared during a presentation at the 25th annual Canadian Telecom Summit in Toronto.

The role of AI in telecom, from network deployment to customer service, was one of the key themes of this year’s conference, which concluded Wednesday.

Cameron McFadyen, the research firm’s vice-president for Western Canada, said satisfaction is lacking among telecom customers about companies’ usage of AI, but there is opportunity to improve.

One of the challenges is rooted in consumers’ skepticism around why AI usage is ramping up. Around four-in-10 believe companies stand to benefit the most from AI in the future, as opposed to the customers themselves.

“(The concern is) in improving their profitability, improving their efficiency, they’ll make the customer experience worse and the consumer will be the one that loses in that transaction,” said McFadyen.

Charlene Gavel, president and CEO of SaskTel, said that as telecom providers increasingly adopt automated tools, they must find a way to meet customers’ needs in a way that creates trust.

“We have very complex systems, very complex processes, very complex customer journeys, so using AI to make that better for the customer is so important, but it has to be done in a way that’s trusted,” she said during a panel discussion.

When it comes to customer service, one of the hopes is that AI can help standardize processes and outcomes, said Rob Bevington, vice-president of technology development at Cogeco.

He said that means customers don’t have to wonder whether waiting on hold to speak to an agent or reaching out through digital channels is the best way to deal with an issue.

“We’re delivering a common experience and creating an equal set of capabilities across all of those channels,” he said.

“I think that removes a lot of friction that customers are experiencing.”

Ipsos’ research shows 43 per cent of consumers are willing to use AI to provide advice or suggestions about telecom services. Just four per cent said they would let AI act on its own to make purchasing decisions.

McFadyen said customers generally still prefer to speak with a human when dealing with a “complex” issue, however many are open to using AI to research plan pricing and help them decide whether to switch providers.

“What the appetite right now is for kind of low-stakes telco experiences, so this is how we can open the door from an AI perspective to get customers comfortable and familiar with these tools,” he said.

AI can help limit service disruptions, but still early days: experts

Experts say AI’s role within telecom operations is only growing behind the scenes — even if the average customer doesn’t realize it.

For the providers themselves, automation is helping to more efficiently manage networks and minimize disruptions to consumers, said Elena Fersman, Ericsson’s vice-president and head of AI innovation and incubation.

“When AI misbehaves, then it’s like, ‘No, we don’t like it.’ But when everything is smooth, then you don’t think about it,” Fersman said in an interview on the sidelines of the conference.

She added that AI tools can enable a “more stable and more optimal” experience for users with fewer outages.

“We use AI to kind of compensate for the imperfections of the network because things happen,” Fersman said.

“The telecom sector benefits a lot from AI algorithms, from machine learning, from intent-based operation, from all the automation techniques. So it becomes more resilient, more reliable, more proactive.”

Manav Gupta, chief technology officer at IBM Canada, said the telecom industry is in the early stages of utilizing AI for that “proactive maintenance” function.

While the goal is to be able to fix problems before a “catastrophic” disruption to a network, that means having a complete set of trustworthy data the AI models can depend on.

And just like commonly used platforms like ChatGPT are still prone to hallucinations due to faulty training data or incorrect assumptions, Gupta said the same can happen with models relied upon by the telecom sector.

“Can you collect all that data across the networks, across the country, for all of the subscribers, for all of the regions, and then use AI in order to figure out … ‘this device is going to go down, you better do some maintenance on that?’ I think they are early in that cycle right now,” he said.

“The worst thing that can happen is without sufficient guardrails, an organization makes a decision thinking that, ‘Oh, this device requires maintenance,’ but turns out it’s the wrong device.”

This report by The Canadian Press was first published May 13, 2026.