Speakers:

Marc Campagna, Chief Executive Officer and Co-Founder – Gaiia (chairperson)
Carlos Anchia, chief AI officer – Resound
Suzy Hays, president and CEO- C Spire
Gene Crusie, CEO- Surf Internet
Matt Larsen, CEO – Vistabeam Internet

There is no shortage of ambition when it comes to artificial intelligence in the communications sector. What remains scarce is the honest account of what it actually takes to move from a promising pilot to something that works in production, at scale, with real customers on the other end. At Metro Connect 2026, four executives at different stages of that journey gathered to share exactly that, the wins, the missteps and the cultural battles that nobody puts in a press release.

The panel, chaired by Marc Campagna, CEO and co-founder of Gaiia, brought together Carlos Anchia, chief AI officer at Resound Networks; Suzy Hays, president and CEO of C Spire; Gene Crusie, CEO of Surf Internet; and Matt Larsen, CEO of Vistabeam Internet.

Campagna set the tone with a striking observation drawn from Gaiia’s own data: whilst 50% of AI agents being deployed today are in software engineering companies, the telecoms industry remains largely untapped territory.

“There’s not enough AI effort being done in our space,” he said, “and we’re trying to push an industry forward.” He added that only 5% of companies successfully move from AI investigation to implemented agents.

The cultural challenge nobody planned for

Before any of the panellists had deployed a single agent, they had all encountered the same obstacle: their own people. The fear of job displacement, it turned out, was not just a communications challenge – it was an operational one, directly slowing adoption.

Crusie described how Surf Internet’s AI governance committee made it their first order of business to address that fear head-on. “We went around with that committee and met with each team, and the first thing the customer care team said was: when do we lose our jobs?” he recalled.

“We told them – we’re not here to replace your job. We’re all about exceptional customer experience.” That reframing, he said, unlocked something. The care team subsequently developed an AI agent that monitors customer satisfaction scores, listens to every support call and tracks competitive dynamics, work that would have been impossible for a human team to do at the same scale or speed.

The outcome was concrete. When the system identified that customers tolerated outages but simply wanted to be kept informed, Surf built an automated circuit record and proactive messaging workflow around that insight. “Our NPS score rose by five in the next quarter,” Crusie said.

Hays described a similar dynamic at C Spire, where the company’s 60-strong developer team had to be brought on a journey rather than simply handed new tools. “I had a great chat with one of our developers who didn’t want AI near us,” she said. “He told me he hasn’t written a line of code since a certain version got so good, but he’s more busy than he’s ever been.”

That developer had since built a churn model in three weeks that delivered direct insight into C Spire’s fibre business. “That’s what I want naturalised across the team,” Hays added. “I really believe that once we get through that initial push, what they’re doing is going to be a lot more fun and a lot more meaningful.”

From prototype to production: where things break

The gap between a working prototype and a production-ready deployment proved to be where most of the hard lessons were learned. Across the panel, the common culprits were data governance, insufficient testing time and an overestimation of what AI actually knows.

Hays shared an anecdote from C Spire’s marketing department, which deployed a team of named AI agents to build a website. The project came in at 75% less cost than a traditional approach and was delivered four to five times faster.

But, it launched with errors. Including, she noted with some amusement, selling the wrong products online. “The important lesson was: it’s not a smart AI,” she said. “AI just knows everything, and it’s believable, and that’s not true. You’ve got to keep your humans in the loop.”

For Larsen at Vistabeam, the challenge was getting enough time from customer care staff to properly test a customer-facing agent before launch.

“We turned on our agent, and then we turned it off,” he admitted. “We’re working through it. The customer service people are not necessarily all in, because the answer to ‘is this going to take my job?’ has to come from actually using the tool and helping make it work.” He noted that adoption varied sharply by department: network operations had embraced AI tools for years, whilst other functions remained harder to move.

Anchia, who joined Resound Networks as chief AI officer six months before the panel, offered a broader perspective on the organisational plumbing required before AI can deliver. “One of the challenges initially was reaching out to every single team to identify what’s on their roadmap for data, how they’re using it, and whether that aligns with how we see our data being used,” he said.

In many cases, the work of deploying AI surfaced underlying data hygiene and governance issues that had previously gone unaddressed. “We find ourselves doing a lot of iteration, going back and providing the right discipline around how we treat our data, the sanitisation of that data.”

Building, buying and measuring what matters

A recurring question throughout the session was how operators should decide what to build themselves versus what to buy or partner on. The panel’s consensus was pragmatic: build where you have a proprietary advantage, buy everywhere else, and be honest about which is which.

“We want to invest our time in customer-facing and proprietary tools that will benefit the user experience,” said Anchia. “If it’s not that, we’re happy to outsource it to a platform that’s going to continue to develop that technology.”

Hays described C Spire’s approach as “buy first, then build on top”, acquiring platforms for scale whilst layering differentiation on top. With operations spanning wireline, wireless and managed IT services across multiple states, scalability was non-negotiable.

When it came to measuring success, Crusie cut through the noise. “At the end of the day, broadband is going to be a commodity product,” he said.

“The one thing that will differentiate us is customer experience.” Surf is now using AI to analyse data across its entire subscriber base – identifying the characteristics of its stickiest customers and using that intelligence to prioritise where it builds next. “What areas look like this area that just performed really well, and how can we build it for about the same cost?” he said. “That’s the platform we’re going to finish building in the next year.”

Hays, meanwhile, set an ambitious benchmark for what success looks like at C Spire following the appointment of a dedicated chief AI officer.

“I hope that every team member and every customer gets the full benefit of AI as we know it today, and as it continues to develop,” she said. “I want us operating about ten times faster than we do now.” She was quick to add a caveat that felt emblematic of the broader conversation: “We also have a very important Chief Information Security Officer. We sell cybersecurity. So yes, but in a secure way.”