Ep 96 – How TELUS turned its data swamp into an AI oasis

Fuel iX is TELUS’s enterprise-grade AI engine launched in 2024 to help companies upgrade their GenAI pilots to production scale in a safe and secure manner. The platform taps into more than 100 LLMs and is used internally by TELUS team members (around 35,000 daily users) and offered to enterprise customers. Internally, employees use it for optimizing models, writing code, generating images, and analyzing complex reports. One standout application is the internal IT support system, which has handled over 70,000 chats and increased self-serve resolution by 49%, enabling 25% cost savings in just 12 months. 

According to Jaime Tatis, building specialized telco-specific LLMs may not be necessary at this stage. While some operators like SK Telecom have invested over $100 million in developing telco-specific models with Anthropic, Tatis believes developing and maintaining industry-specific LLMs is extremely costly and resource-intensive, with ROI that may not justify the spend. Instead, TELUS focuses on adapting existing general-purpose models that can be fine-tuned with specific datasets to meet telco needs without building from scratch. As Danielle Rios from TelcoDR notes, the rapid advancement of AI technology makes flexibility in choosing LLMs essential.

TELUS partnered with Google Cloud in 2020 to radically transform its data infrastructure rather than simply lifting and shifting legacy systems. The company deprecated over 30% of obsolete datasets from what it called its “data swamp,” built new data pipelines using modern SRE principles, and consolidated everything into a single enterprise data hub with strong privacy and security by design. This transformation broke down data silos and democratized access to curated information across the organization. Read more about how TELUS worked with Google Cloud and Accenture to modernize its data stack and create AI-ready foundations.

TELUS created and won awards for the Neo Assistant, an AI-powered solution that significantly improved network technician efficiency. Using machine learning, it enhanced forecasting accuracy for network events from 69% to 89%, helping technicians work smarter and make the right fix the first time. The GenAI-powered device-accessible solution guides technicians through tasks, achieving 75% adoption by field workers. Combined with other AI solutions, TELUS reduced network operating costs by over $20 million (some reports cite $15 million specifically for network operations). For more inspiration on AI applications, check out DR’s blog about 10 examples of how telcos are using AI.

TELUS developed an AI-powered translation copilot to help team members communicate in both French and English, Canada’s official languages. The tool was trained to understand company-specific lingo and tone, achieving an 87.5% accuracy rate compared to professional human translators in blind testing. Beyond matching quality, the solution provides instantaneous translation rather than the multi-day turnaround required for professional services, dramatically accelerating business communications. This demonstrates how TELUS’s commitment to responsible data practices enables AI applications that deliver immediate business value.

As DR emphasizes in her Telco in 20 takeaway, making AI work depends entirely on data quality. Many telcos sit on siloed, obsolete information trapped in legacy vendors’ databases. It isn’t clean or structured enough for AI applications. You can’t just lift and shift that data to the public cloud and expect miracles—radical data transformation is required. TELUS’s experience shows that while this transformation takes time and effort (along with hyperscaler support), the payoff is huge. If you’re ready to transform your data infrastructure for AI success, set up a meeting with TelcoDR or reach out to DR on LinkedIn or X @TelcoDR.