What if most companies are doing AI wrong? That's the question animating Bell Canada's approach to customer experience. While enterprises race to deploy chatbots that deflect calls and answer simple questions, Canada's largest telecommunications company sees a missed opportunity. And now, it’s betting on a more ambitious path forward.
“We’ve all been seduced by the large language models and the conversationalist nature of AI as opposed to its true potential on driving actual enterprise outcomes,” says Lukas Lhotsky, who leads Bell’s platform transformation (and its subsidiary Ateko). “We need to move beyond a paradigm of simple conversation toward a paradigm of getting things done: a paradigm of action.”
Rather than treating AI as a frontline filter, Bell is embedding AI agents throughout its entire operations to proactively resolve issues, orchestrate and automate complex workflows, and augment the human agent experience.
For Rohit Batra, vice president and general manager of industry products at ServiceNow, Bell's approach mirrors industry trends for customer experience AI. “The future isn’t just better chatbots. It’s about AI that can actually finish the work—across billing, network ops, field service—to resolve customer issues, not just escalate and route them around.”
When Bell first started investing in AI, the company wanted to deploy it across 14 platforms, all connected through APIs, says Lhotsky. “And I thought: That sort of complexity means they will never achieve their potential.”
Instead, Lhotsky advocated for a different approach: "building backup cameras, not iPhones."
A backup camera is tailor-made for one task, and it does it exceptionally well. The power of an iPhone, on the other hand, is more about breadth. It’s pretty good for watching videos, listening to music, and browsing the web—but not as good as a 75-inch TV, a high-end audio system, or a laptop. Rather than designing AI agents that can complete multiple tasks, Bell focuses on narrowly scoped AI agents designed to excel at specific tasks across an integrated platform.
It’s a choice that sets Bell apart from the rest of the market, says Batra. "Most telcos are still in pilot mode, bolting on AI tools that sit alongside disconnected systems that were never designed to work together. Bell is doing the opposite. It’s deploying AI that finishes work across the front, middle, and back office on one common platform. That’s the only way you get from building demos to delivering outcomes.”
One of Bell's more significant shifts is moving from reactive to proactive customer communication. When a fiber cut disrupts service, for example, the system automatically creates a ticket, identifies affected customers, and pushes out notifications before they pick up the phone.
"People are getting text messages, and they don't have to call," Lhotsky says. “‘Hey, there's a fiber cut. We're aware of it. We're actively working on it. We expect a resolution ETA of X. And, by the way, you now have free cellphone data if you want to use it for the next few hours.’”
It's a different way of thinking about call deflection. Rather than rerouting customers to automated systems when they’re seeking help, Bell aims to eliminate the need for them to seek help in the first place.
Bell transcribes every customer call in real time. But rather than using that capability to replace human agents, the telecom company feeds those transcriptions to AI agents that assist as conversations unfold.
"While the human agent is being an empathetic listener, you can have a partner AI agent digging through recent transactions and validating things," Lhotsky explains. The goal is to eliminate the barriers that pose an inconvenience for customers: the "give me a moment" delays while AI agents search through multiple systems for answers.
Bell's approach reflects what Batra considers fundamental: using AI to support human decision-making, not disrupt it. “AI isn't replacing human agents; it's giving them their jobs back,” says Batra.
"When AI handles the repetitive work like billing updates, provisioning, and quote creation, teams get time back for what requires real judgment: complex cases, diagnosing a network issue, building relationships. They can resolve issues faster and with more confidence because they’re not drowning in system work."
AI handles what machines do well: gathering relevant information, triaging cases, surfacing resolution steps, and automatically generating wrap-up notes. Humans handle what humans do well: listening, empathizing, and exercising judgment in ambiguous situations.
The results speak for themselves. AI-driven coaching has helped human agents handle 180,000 calls, more than 500,000 customers have been assisted via AI-powered Virtual Repair, and customers have contacted agents 2.6 million fewer times in 2025 than in 2024.
For companies looking to follow a similar path, Lhotsky encourages patience and focus. "Think small at the beginning," he advises. "People become overexcited and lose focus on the business outcome they're trying to achieve."
His recommendation is to start with narrowly defined AI use cases where the data already exists and the workflows are well understood. Build momentum through quick wins. Only then expand to more ambitious applications.
"Where are the really common-sense use cases where we can launch a mission-specific AI, where the data set is there, the underlying workflows are appropriate?" he says. "Go build the momentum and the use cases there. And only then can you start to move upward.”
Two years into its transformation, Bell estimates another two to three years of work ahead. But Lhotsky sees a clear destination: a future where the technology becomes so embedded that business teams can launch new initiatives through conversation with an AI agent.
"I really hope we get to a day where the platform is so material that it's immaterial technologically," he says, "and it's only material from a business point of view."
Find out how ServiceNow can help put AI agents to work for people.
