TELUS Digital at Scale: What 2 Trillion AI Tokens and 6,000 Custom Assistants Reveal About Enterprise AI Deployment

Executive Summary

• TELUS Digital processed 2 trillion AI model tokens in 2025 through its Fuel iX platform — the largest internal AI usage volume in the corporate-tools corpus. Scale alone does not equal ROI, but it is a meaningful signal that they are past experimentation.
• The headline “$100M+ value” claim appears in the MWC 2026 session title and is not broken down by use case, time period, or business unit in any public document. Treat as unverified headline framing, not an audited figure.
• The 6,000+ custom GenAI assistants across 50,000+ TELUS employees is the more instructive data point: it shows that no-code AI customization at enterprise scale is achievable. This figure is from MWC 2025 context (the internal TELUS parent-company case) and was the flagship example used to market Fuel iX externally.
• TELUS Digital is a BPO and digital services arm serving 30+ global telecom brands — not a standard internal IT deployment. Their scale is structurally atypical for most mid-market enterprises. The applicable lesson is the sequencing model, not the volume.
• The platform architecture (multi-model MLOps, per-token cost tracking, no-code assistant creation, ISO Privacy by Design certification) maps to decisions CIOs face now: build a governed AI platform or buy point tools.

What TELUS Digital Actually Deployed

The Fuel iX Platform

Fuel iX is TELUS Digital’s proprietary GenAI engine, built to serve both internal TELUS operations and external telecom clients. It is not a single model — it provides access to 20+ LLMs from multiple vendors through a central control plane. Key architectural decisions embedded in the platform:

• Per-token cost tracking by use case. Every AI call is logged against a specific business process. This is what separates a platform from a set of point tools — it makes cost accountability possible.
• Multi-model MLOps. No lock-in to a single foundation model. Teams can route different tasks to different models based on cost, latency, and capability.
• Central data governance. Role-based access controls, encryption enforcement, and privacy policy applied at the platform layer — not patched in per application. The platform received the world’s first ISO Privacy by Design certification for a GenAI application.
• No-code assistant creation. Employees without technical expertise can build custom assistants. This is what produced 6,000+ custom tools across 50,000+ employees without a centralized IT bottleneck.

Internal TELUS Deployment (The 6,000 Assistants Case)

This is the flagship internal case that TELUS Digital uses to demonstrate Fuel iX to external telecom clients. The figures are from the MWC 2025 presentation:

• 6,000+ custom GenAI employee assistants created by TELUS employees (no-code builds)
• 50,000+ global employees covered
• 500,000+ hours saved across the deployment
• ~40 minutes saved per AI-assisted interaction
• 35,000 employees actively using Fuel iX Copilots at product launch (earlier baseline)

The progression from 35,000 active users at launch to 50,000+ covered employees with 6,000+ custom tools illustrates the compounding effect of no-code access: deployment scales faster than IT can provision tools centrally.

Cross-reference: The Anthropic Agentic Coding Trends 2026 report (research/01-ai-native-landscape/anthropic-agentic-coding-trends-2026.md) cites TELUS with a slightly different framing — 13,000 custom AI solutions, 30% faster engineering code shipping, 500,000 hours saved. The discrepancy between 6,000 (Fuel iX / MWC 2025 framing) and 13,000 (Anthropic framing) likely reflects different time periods or different counting methodologies (Fuel iX assistants vs. all AI solutions including externally-facing ones). Neither figure is independently audited. Use the 500,000 hours / 40 minutes per interaction as the more verifiable operational metric.

External Telecom Client Deployments (2026)

The MWC 2026 press release focuses on Fuel iX as a product for external telecom operators, not just internal TELUS use. Three solution categories were demonstrated:

1. Fuel iX Agent Trainer AI-powered voice and chat simulation for contact center agent onboarding. Claims faster ramp time and cost reduction vs. live-call training. No specific time or cost figures published in press materials.

2. Fuel iX Fortify Automated testing for AI safety and security: toxicity detection, PII leakage scanning, guardrail validation, red-teaming. This is a separate productized capability addressing an enterprise compliance gap — most organizations deploying GenAI do not have systematic testing for PII exposure in model outputs.

3. Network Design Services AI-driven network planning and optimization for telecom operators moving from legacy OSS/BSS stacks to cloud-native architectures. Integrated with TM Forum Open Digital Architecture standard.

The $100M+ Value Claim

The $100M+ figure is the session title of TELUS Digital’s MWC 2026 presentation: “AI transformation in telecom: Unlocking $100M+ value with innovative use cases.” No public document breaks this down by use case, customer, or time period. It is not an audited or verified figure. It functions as positioning for the external services business — TELUS Digital is selling Fuel iX to other telecom operators, and $100M+ is the promise, not a reported result.

Bottom line on credibility: The $100M+ headline is MEDIUM at best as a directional claim for what is achievable across a large telecom operator’s AI program. It cannot be cited as a specific outcome.

The Industry Context: Where Telecom AI Stands

TELUS Digital cites NVIDIA’s 2025 report on telecom AI priorities:

• 44% of telecom operators prioritize customer experience (CX) optimization as their top AI investment
• 40% focus on network planning and operations

The framing of the MWC 2026 session is “escaping Pilot Purgatory” — the pattern where organizations deploy many pilots but cannot scale to production. This is the same pattern documented across the corpus in supply chain AI (Blue Yonder, Carrefour) and QSR (Papa Johns). TELUS Digital’s argument is that platform infrastructure (Fuel iX) is what bridges pilot to production, not better individual use cases.

This is a vendor argument with a direct commercial interest, but the underlying diagnosis is consistent with what enterprise deployment evidence shows: companies that built governed AI platforms before deploying use cases (Carrefour with SymphonyAI, TELUS with Fuel iX) achieved more durable scale than those that procured point tools per use case.

Key Data Points

What This Means for Your Organization

The 2T token volume and 6,000 custom assistants are not directly applicable benchmarks for a 1,000–10,000 person company. TELUS Digital operates at telecom infrastructure scale with a dedicated AI platform team and 20+ years of BPO delivery infrastructure. The numbers are useful for one purpose: establishing that enterprise-wide no-code AI deployment is achievable in practice, not just in theory.

The more applicable lesson is the platform architecture decision. TELUS did not deploy 6,000 AI tools by having IT build 6,000 tools. They built a governed platform with central data controls and no-code access, then let employees create tools within guardrails. For a 1,000–10,000 person company, this translates to a concrete sequence: define the governance layer (data access policies, approved models, cost tracking, PII guardrails) before enabling broad self-service. Organizations that skip the governance layer and go directly to broad access generate shadow AI — TELUS Digital’s own 2025 survey documented this problem at other enterprises.

The Fuel iX Fortify capability is worth noting separately. Automated testing for PII leakage and toxicity in AI outputs is a gap in most enterprise AI programs. Most organizations have not built systematic validation for what their models output under adversarial prompts. This is a specific action item: before scaling any GenAI deployment, establish automated output testing, not just prompt-input controls.

For a CIO considering enterprise-wide deployment: the TELUS case sets the operational ceiling for what an AI-native platform organization can achieve. Your deployment will look different — narrower scope, tighter governance, fewer custom tools — but the sequencing logic applies. Platform governance first, use-case expansion second.

Questions about enterprise AI deployment sequencing or governance layer design: brandon@brandonsneider.com

Sources

1. TELUS Digital Press Release — MWC 2026 (Feb 24, 2026, PRNewswire)

   • URL: https://www.prnewswire.com/news-releases/telus-digital-showcases-ai-transformation-in-telecom-unlocking-value-with-innovative-use-cases-at-mobile-world-congress-2026-302695271.html
   • Type: Company-published press release
   • Credibility: MEDIUM — marketing framing, no third-party verification, Bret Kinsella quotes are direct attribution
   • Temporal tier: TIER 1 (Feb 2026)

2. Fuel iX MWC 2025 Case Study / fuelix.ai

   • URL: https://www.fuelix.ai/post/telus-digital-experience-launches-fuel-ix-copilots-an-enterprise-safe-generative-ai-solution-for-employee-productivity-and-knowledge-search
   • Type: Company-published product launch announcement
   • Credibility: MEDIUM — self-reported metrics, direct attribution to TELUS deployment, no control group
   • Temporal tier: TIER 2 (2024–2025 deployment data)

3. Anthropic Agentic Coding Trends 2026 Report (Apr 2026) — TELUS cross-reference

   • Internal research file: research/01-ai-native-landscape/anthropic-agentic-coding-trends-2026.md
   • Note: Reports 13,000 custom AI solutions (vs. 6,000 in Fuel iX framing) — likely different time period or scope definition
   • Credibility: LOW — vendor case study published by Anthropic, no control group. These case studies are vendor-published and represent selected wins with no control group and no independent verification. Cross-reference against: METR RCT (experienced developers 19% slower), CMU study (40.7% code complexity increase), Atlan 200-deployment analysis (median +159.8% ROI requires workflow redesign first).

4. NVIDIA 2025 Telecom AI Report (cited in TELUS Digital press release)

   • Cited as: NVIDIA report on telecom operator AI investment priorities
   • Credibility: MEDIUM — vendor-commissioned research with direct commercial interest in telecom AI investment

5. TechNexus / hipther.com — secondary aggregators republishing PRNewswire content

   • Credibility: LOW — no independent reporting, direct republication of press release

Brandon Sneider | brandon@brandonsneider.com April 2026