AI "accent masking" technology that changes how call center workers sound is drawing criticism from unions in Canada. The technology raises uncomfortable questions about authenticity, worker dignity, and who benefits from making people sound "less foreign."
Telus Digital has partnered with an AI-powered speech enhancement company to "soften accents" at overseas call centers. The technology modifies workers' voices in real-time, making them sound more like native English speakers to North American customers.
Let me be blunt: this technology is technically impressive and ethically horrifying.
Companies are essentially asking workers to wear digital masks so customers don't know they're speaking to someone overseas. The message is clear: your actual voice isn't acceptable. We need to hide who you are to satisfy customers who don't want to talk to you.
Labor representatives have expressed significant backlash against the technology. Unions view accent modification tools as problematic for workers in both domestic and international call center operations. The controversy centers on what this technology says about bias, authenticity, and who has to code-switch to keep their job.
From a pure engineering perspective, this is genuinely sophisticated work. Real-time voice modification that preserves meaning while altering accent markers requires serious AI capabilities. But just because we can build something doesn't mean we should.
The business case is obvious: some customers hang up or demand different representatives when they detect foreign accents. Accent bias is real, and it costs companies money in repeat calls and customer satisfaction scores. The technology lets companies keep overseas workers while hiding the thing customers object to.
But solving customer bias by masking worker identity is choosing the wrong problem to fix. If your customers have issues with accented English, maybe the solution is addressing customer bias, not technologically erasing worker identity.
The workers themselves haven't been prominently featured in discussions about this technology, which is telling. The power dynamics suggest the latter.




