Telus Deploys AI to Modify Call-Agent Accents

Telus, one of Canada's largest telecommunications companies, is deploying AI-powered accent-modification technology in its call centers to alter how agents sound to customers in real time. The move places Telus at the forefront of a growing — and controversial — trend in customer service automation that raises profound questions about bias, identity, and the future of global outsourcing.

The technology works by processing an agent's speech in milliseconds, converting their natural accent into one that more closely resembles a North American or neutral English accent. While Telus frames the initiative as a way to improve customer satisfaction and reduce miscommunication, critics argue it papers over deeper issues of accent-based discrimination rather than addressing them.

Key Facts at a Glance

• What: Telus is using real-time AI accent-modification software in its customer service operations
• Why: To reduce communication friction, improve customer satisfaction scores, and lower call handling times
• How: AI models process speech in real time, modifying accent characteristics while preserving the speaker's words and tone
• Who's affected: Thousands of call-center agents, primarily based in the Philippines, India, and Central America
• Industry context: Telus joins a growing list of companies exploring accent AI, alongside vendors like Sanas AI and Krisp
• Controversy: Critics say the technology reinforces accent-based bias instead of combating it

How the Accent-Modification Technology Works

The underlying technology relies on deep learning models trained on thousands of hours of speech data across multiple accents and dialects. These models learn the phonetic patterns, intonation curves, and prosodic features that distinguish one accent from another.

During a live call, the system captures the agent's voice, processes it through a neural network, and outputs a modified version — all within approximately 20 to 50 milliseconds. This near-zero latency is critical for maintaining natural conversation flow.

Unlike simple voice filters or pitch-shifting tools, modern accent AI preserves the speaker's unique vocal qualities, emotional inflection, and word choice. The system only modifies accent-specific characteristics such as vowel pronunciation, consonant articulation, and rhythmic patterns. The result is a voice that still sounds human and natural but carries a different regional accent.

Companies like Sanas AI, a Silicon Valley startup that has raised over $50 million in funding, have pioneered this approach. Sanas reports that its technology can reduce average call handling time by up to 28% and improve customer satisfaction scores by as much as 23%. Telus has not disclosed which specific vendor or in-house solution it uses, but the operational principles remain similar.

Why Telus Is Making This Move Now

Telus operates one of the world's largest business process outsourcing (BPO) networks through its subsidiary Telus Digital (formerly Telus International). The company employs over 75,000 team members across more than 30 countries, with significant operations in the Philippines, India, Guatemala, and El Salvador.

Customer service has long been a domain where accent bias creates measurable business problems. Research from the University of Manchester and other institutions has shown that listeners often rate speakers with non-native accents as less credible, less competent, and harder to understand — even when the content is identical.

For Telus, this translates directly into business metrics:

• Higher call abandonment rates when customers perceive communication difficulty
• Longer average handle times due to repeated clarifications
• Lower first-call resolution rates as misunderstandings lead to callbacks
• Reduced Net Promoter Scores (NPS) tied to perceived service quality
• Increased agent turnover as workers face frustration and even verbal abuse from callers

By deploying accent AI, Telus aims to neutralize these friction points without relocating operations to higher-cost English-speaking markets. The economic calculus is straightforward: accent modification costs a fraction of what it would take to hire native-accent agents in the U.S. or Canada, where call-center wages can run $18 to $25 per hour compared to $3 to $7 per hour in offshore locations.

The Ethical Debate Intensifies

Not everyone views this technology as progress. Labor advocates, linguists, and diversity experts have raised sharp objections, arguing that accent modification sends a troubling message: that certain accents are problems to be fixed.

'This technology essentially tells workers that their natural voice isn't good enough,' said one labor rights researcher who studies BPO industry practices. 'Instead of training customers to be more patient or addressing bias, companies are asking workers to digitally mask their identity.'

The parallels to other forms of identity erasure in the workplace are hard to ignore. Critics compare accent AI to policies that once required workers to anglicize their names or suppress cultural markers to fit corporate norms.

However, supporters of the technology offer a different perspective. They argue that accent modification is a pragmatic solution that actually benefits agents by:

• Reducing verbal abuse and discriminatory behavior from callers
• Lowering stress levels during interactions
• Improving performance metrics, which can lead to better compensation
• Preserving jobs that might otherwise be automated entirely or moved onshore

Sanas AI CEO Maxim Serebryakov has previously stated that the company's goal is to 'make communication smoother for everyone' and that agents themselves often welcome the technology because it reduces the hostility they face daily.

How This Fits Into the Broader AI Landscape

Telus's accent AI deployment is part of a much larger transformation sweeping the $350 billion global BPO industry. Call centers have become one of the most aggressive adopters of AI technology, moving far beyond simple chatbots into sophisticated real-time assistance tools.

The current generation of contact center AI includes:

• Real-time accent modification (Sanas AI, Krisp)
• AI-powered agent assist tools that suggest responses during live calls (Google CCAI, Amazon Connect)
• Automated quality assurance that monitors 100% of calls instead of random samples (Observe.AI, CallMiner)
• Predictive routing that matches callers with optimal agents based on AI analysis
• Generative AI summarization that eliminates post-call documentation work

Compared to fully autonomous AI agents — like those being developed by Sierra AI, Bland AI, and others — accent modification represents a middle-ground approach. Rather than replacing human agents entirely, it augments their capabilities while keeping humans in the loop.

This is significant because fully autonomous voice AI still struggles with complex, emotionally charged, or ambiguous customer interactions. McKinsey estimates that while AI could automate up to 30% of contact center tasks by 2026, the majority of interactions will still require human judgment for the foreseeable future.

What This Means for the Industry

Telus's adoption of accent AI signals a broader shift in how companies think about the economics and ethics of global outsourcing. If the technology proves effective at scale, it could have several ripple effects.

For BPO companies, accent AI removes one of the last competitive advantages that onshore call centers held over offshore operations. If a Filipino agent can sound indistinguishable from a Canadian agent, the cost advantage of offshore operations becomes even more compelling.

For workers, the implications are mixed. On one hand, accent AI could protect offshore jobs that might otherwise be lost to automation or reshoring. On the other hand, it introduces a new form of digital surveillance and identity modification that workers may not have meaningful power to refuse.

For consumers, the change may be largely invisible. Most callers are unlikely to know or notice that the voice they hear has been AI-modified. This raises questions about transparency — should companies disclose when accent AI is in use?

For regulators, accent modification adds another item to the growing list of AI applications that may require governance frameworks. The EU AI Act, which takes effect in stages through 2026, could potentially classify real-time voice modification as a system requiring transparency obligations.

Looking Ahead: The Future of Voice AI in Customer Service

The accent-modification market is still in its early stages, but growth projections are steep. Analysts estimate the broader conversational AI market will reach $32 billion by 2028, with real-time voice processing tools capturing an increasing share.

Several developments are likely in the near term:

Within 12 months, expect more major BPO providers — including Concentrix, Foundever, and Wipro — to announce similar accent AI deployments. The competitive pressure to match Telus's efficiency gains will be intense.

Within 2 to 3 years, accent AI will likely become bidirectional, modifying not just agent accents but also translating between languages in real time. Meta's SeamlessM4T and similar multilingual models are already laying the groundwork for this capability.

Within 5 years, the line between accent modification and full voice synthesis may blur entirely. Advances in voice cloning and text-to-speech technology could enable companies to create entirely synthetic agent voices that are indistinguishable from real humans.

Telus's decision to deploy accent AI today is a bet that the technology's benefits outweigh its risks. Whether that bet pays off — for the company, its workers, and its customers — will depend not just on the technology itself, but on how transparently and ethically it is implemented. The conversation about who gets to decide what a 'normal' accent sounds like is only just beginning.