AT&T Has 500 Agentic AI Workflows Running and Plans to Triple That Next Year

Quick Facts

• AT&T runs nearly 500 agentic workflows in production and plans to triple that number in 2027.
• The company processes 45 billion AI tokens daily and reports a fivefold cash return on AI investment.
• Agentic automation has cut provisioning time for dedicated internet connections in half and saved over 100 million pounds of carbon emissions from fleet optimization.

AT&T has moved past generative AI experimentation and is now redesigning core business operations around autonomous agents. The company runs nearly 500 agentic workflows in production and plans to triple that count next year, according to Luke Corbin, assistant vice president of digital transformation and AI at AT&T Business.

The scale is significant. AT&T's agentic platform handles more than 1 million transactions monthly. Its Ask AT&T and Ask AT&T Workflows platforms serve 100,000 users and have logged 750 million API calls into production systems.

The Financial Case

Andy Markus, AT&T's chief data and AI officer, said the company hit a fivefold cash return on AI investment last year. "We've hit the flywheel," Markus said. The company reports more than a twofold in-year ROI on generative AI initiatives, along with $17 million in annual operational savings and an 80% reduction in fraud.

Customer-facing agents deliver 33% faster resolution times. Corbin said production AI investments are achieving fivefold returns within a year, with success measured across labor savings, revenue generation, operating costs, and customer satisfaction.

How the Architecture Works

AT&T keeps costs in check by routing work to smaller language models when full-scale models are unnecessary. "When the super agents call subagents, a lot of them are small language models," the company said, adding that this approach is saving 90% of the cost compared to routing every token through frontier models. The infrastructure runs on Azure.

Markus described the design philosophy plainly: "We're taking the bigger complex problems, breaking them down into smaller pieces and then we're creating agents to be great at each piece, bringing those together in an agentic workflow to solve the overall solution in a much more accurate way."

The Business Agent Foundry

AT&T established the Business Agent Foundry in early 2025 to drive this work. The group embeds with individual business units, maps existing processes using industry frameworks, and identifies where manual steps, organizational handoffs, and legacy application dependencies can be cut or automated.

Rather than collecting ad hoc automation ideas from employees, the foundry takes a top-down view of business processes. Corbin said responsibility for AI-driven transformation should eventually be distributed across the organization rather than concentrated in a central team.

Specific Operations Rebuilt

Several concrete use cases show how deep the changes go. Changes to fiber installation orders previously required employees to pull tickets manually, update records, and return them to the workflow. Agents now handle that process. AT&T also cut in half the time needed to provision dedicated internet connections in buildings where fiber infrastructure already exists.

In network operations, three agents now handle fault detection and resolution. The first correlates telemetry data, checks change logs, and opens a ticket. The second suggests a resolution and writes new code. The third compiles artifacts and generates a summary for the engineer, who stays in the loop throughout.

AT&T operates more than 50,000 vehicles and faces over 700 million possible routes on any given day. An AI dispatch optimization application has saved more than 100 million pounds of emissions since launch by reducing miles driven.

For fraud prevention, AT&T moved beyond flagging suspicious calls. Agents now engage directly with unknown callers using voice-to-voice capabilities to determine whether a call is legitimate before it reaches a customer or employee.

Legal Operations

In January 2026, AT&T launched LegalEdge, an in-house legal team of attorneys, paralegals, technologists, and e-discovery specialists built around AI-assisted work. The group handled nearly 100 legal matters in its first six months. AT&T is working with OpenAI as a flagship in-house legal design partner on the initiative.

Corbin said AT&T is now shifting focus from improving existing processes to using agents to launch new services, enter new markets, and open new sales channels. For technology executives watching from the outside, AT&T's model is a blueprint for what agentic deployment looks like at scale.