When you call TELUS customer service, you may be speaking with an agent in the Philippines or El Salvador. That has been true for years and, as a business decision, is largely unremarkable. What is new, and what deserves a direct answer from Canada’s third-largest telecom, is this: the voice you hear has been run through artificial intelligence software that alters how the agent sounds before the audio reaches your ear.
TELUS Digital confirmed in May 2026 that it uses technology from a company called Tomato.ai, which applies speech-to-speech AI models in real time to reduce what the company describes as accent-related friction. The software encodes the agent’s voice, modifies acoustic and pronunciation features, and sends the altered audio stream to the customer. The agent’s words are unchanged. Their voice is not.
What TELUS Says, and What It Does Not Say
The company frames the tool as a clarity and comprehension aid. That is a reasonable thing to want. Anyone who has ended a support call frustrated by miscommunication understands why businesses look for fixes. TELUS has also said it will move toward clearer disclosure, stating it plans to inform customers when AI-assisted technologies are in use on calls, including tools designed to improve voice clarity. That commitment, made after the story broke publicly, is not the same as having disclosed the practice beforehand. Customers were never told while the technology was already deployed.
Unifor telecommunications director Roch Leblanc put the concern plainly: using AI to alter offshore agents’ voices could mislead Canadians into thinking they were speaking with Canadian-based employees. Rogers and Bell have both said they will not follow TELUS’s path here. That is a meaningful signal from two direct competitors who looked at the same tool and passed.
A Deception Problem, Not Just a Technology Problem
The argument for accent-neutralisation software tends to land on worker welfare. Agents in offshore centres do face accent-based hostility from callers, and reducing that friction has genuine human value. But the customer on the receiving end of that call is not being protected by the software. They are being managed by it. There is a difference between improving audio quality and altering the perceived identity of the person speaking. The first is a technical fix. The second shapes what the customer believes about who they are talking to.
Canada has no federal rule requiring disclosure when AI modifies the voice of a human agent in real time. That gap matters more as the technology spreads. If TELUS is using Tomato.ai today, other companies are evaluating it. Without a disclosure standard, the default will be silence, because silence costs nothing and disclosure might prompt awkward questions about offshore staffing levels.
Alberta’s AI Economy Needs an Honest Foundation
Alberta is actively positioning itself as a destination for AI investment. Data centres are going up across the province. The provincial government has spoken openly about capturing a share of the AI buildout happening across North America. That ambition is worth pursuing. But an AI economy built without accountability rules is a liability, not an asset.
The TELUS situation illustrates exactly where the gap shows up. The technology was deployed, customers interacted with it for months, and no one was told. A simple disclosure requirement, something like a brief statement at the start of a call that AI tools may be in use to improve audio quality, would cost almost nothing to implement. It would also establish a norm that consumers can calibrate their expectations against. Alberta is already building AI infrastructure faster than Ottawa is building policy, and that gap will widen. Provincial leadership on disclosure standards, even as a model for federal regulators, would be a concrete contribution to a debate that is just beginning.
The Test for What Comes Next
TELUS’s pledge to add disclosure messaging is a start. It came after public pressure, not before. The Canadian Radio-television and Telecommunications Commission and the federal government have been urged by labour groups to require disclosure as a condition of use. Whether those calls lead anywhere depends on whether regulators treat consumer trust as a policy variable or an afterthought. Alberta’s industries are moving fast on AI deployment, and the province has every reason to want the rules settled early rather than retrofitted after the damage is done.
The technology will keep improving. The voices will sound more natural. The gap between a human agent and an AI-mediated one will narrow. That makes the question of disclosure more urgent, not less, because a standard set now, while the public is paying attention, is worth far more than a regulation drafted after trust has already eroded.
If a company modifies how a human sounds before the call reaches you, do you have a right to know?
