AT&T’s Deepak Sharma on Why AI that Feels Like Magic is AI that Works

AT&T serves millions of customers every day across retail, contact centers, and digital channels. Making those experiences seamless is no small task. Deepak Sharma, Head of Retail Technology, Contact Center Platforms, and GenAI Product & Engineering at AT&T, joined AICX Innovators to share how his team is tackling this transformation.

Here are the biggest lessons from his approach.

• Breaking Down Omnichannel Silos

For decades, contact centers, retail, and digital platforms were built independently. That disconnect between online, store, or phone call leaves a customer back at the starting line for each new interaction.

Deepak’s mission is to unify these systems into a truly seamless omnichannel experience. The goal isn’t just efficiency, but trust, loyalty, and brand consistency at AT&T’s scale.

• Building AI-Ready Foundations

World-class CX can’t sit on top of fragmented legacy infrastructure. AT&T began by modernizing its core systems and data architecture, creating a unified orchestration layer to share context in real time.

Deepak emphasized a two-lane approach:

1. Fast Lane: Quick wins like call summarization, guided workflows, and smart routing to deliver immediate value.

2. Core Lane: Longer-term modernization of data pipelines, semantic layers, and orchestration to support AI natively at scale.

• Bringing AI Into the Customer Journey

With this foundation, AI is stitching together customer experiences that were once fragmented:

• Starting a transaction online, then seamlessly resuming in-store or via an agent.
• Digital avatars acting as lifelike assistants that guide customers through product selection and hand off to humans when needed.
• Proactive personalization that anticipates needs, rather than reacting after problems arise.

“Omnichannel isn’t about connecting systems—it’s about delivering a seamless human experience at scale.”

• Overcoming AI Adoption Challenges

Deepak highlighted the top three blockers to enterprise AI adoption:

1. Data & Infrastructure: AI can’t work without clean, connected, accessible data.

2. Expectation Management: Aligning stakeholders on what AI can do now vs. later.

3. Change Management: Involving frontline teams from the start so adoption feels natural, not forced.

• Making AI Invisible (and Impactful)

AT&T’s north star is making AI so seamless that employees and customers don’t even notice it—it just makes experiences faster, smarter, and easier.

Examples like agent assist show how adoption happens naturally when tools reduce effort, remove friction, and improve outcomes. As Deepak put it:

“The best change is the one people don’t even realize they’re going through.”

• The Role of Partners

System integrators and consulting firms still matter, but their role is evolving. Instead of multi-year ERP rollouts, the new mandate is value orchestration—bringing AI, data, and business context together quickly using modern platforms and pre-built connectors.

• AI at Big Data Scale

With hundreds of millions of customer interactions each year, AT&T can’t rely on massive general-purpose models alone. Instead, Deepak’s team is building a hybrid AI architecture:

Large models for complex tasks.

Smaller, domain-specific models for high-volume, real-time use cases like summarization, routing, and intent detection.

A mix of open-source and fine-tuned models to balance performance, cost, and latency.

Key Takeaways for CX Leaders

• Fix the foundation first: Modernize core systems to enable AI at scale.
• Adopt a two-lane strategy: Deliver quick wins while investing in long-term transformation.
• Prioritize customer value: Use AI to eliminate friction and create continuity across channels.
• Make AI invisible: Design experiences so seamless they don’t feel like “using AI.”
• Rethink partners: Seek accelerators, not just implementers.
• Architect for scale: Balance big models with smaller, efficient ones tuned for CX.

At Level AI, we share this philosophy: the future of CX requires AI that works hand in hand with humans, built on clean foundations and designed to deliver seamless, invisible value at scale. Hear the full podcast here: