AT&T's AI Overhaul: Orchestrating Intelligence at 45 Billion Tokens a Day

MOUNTAIN VIEW, CA – July 14, 2026 – In a move that signals a new phase of enterprise artificial intelligence, AT&T has begun deploying H2O AI Super Agent™ to power its vast and complex AI initiatives. The telecommunications giant, which processes a staggering 45 billion tokens daily across its operations, is embracing a future where AI isn't just a tool for isolated tasks, but a fully orchestrated, autonomous system driving core business functions from customer experience to fraud prevention.

This partnership with agentic AI specialist H2O.ai represents a significant leap beyond the current generation of AI applications. Instead of deploying individual models or simple chatbots, AT&T is implementing a sophisticated 'orchestrator' designed to manage a fleet of specialized AI agents, streamline workflows, and make intelligent decisions at a scale few companies can rival. The integration will enhance the internal 'Ask AT&T' platform and deploy autonomous capabilities across the enterprise, marking a pivotal moment in the industrialization of AI.

The Dawn of the AI Orchestrator

The central challenge for large enterprises like AT&T is no longer just building AI models, but managing them. With thousands of models running across different departments, the risk of creating disconnected, inefficient, and ungovernable 'agent sprawl' is high. The collaboration with H2O.ai tackles this head-on by introducing a multi-agent architecture that acts as a central nervous system for AI operations.

Unlike single-purpose agents, the H2O AI Super Agent functions as an orchestrator. It can take a high-level objective—such as investigating a complex fraud pattern or optimizing field operations—and decompose it into smaller, structured sub-tasks. It then deploys a variety of specialized agents in parallel, including deep research agents, predictive engines, and fine-tuned small language models (SLMs), to execute these tasks. The system dynamically adapts its workflow based on incoming data and context, effectively creating an autonomous problem-solving engine.

“At AT&T, we operate AI systems at enormous scale, processing about 45 billion tokens per day across complex environments,” said Andy Markus, Chief Data Officer at AT&T. “As our AI initiatives expand, we need to reduce operational complexity, improve reasoning across separate systems, and deploy lower-latency, cost-efficient AI with enterprise-grade governance. H2O AI Super Agent helps us orchestrate predictive AI and autonomous agents in a scalable architecture built for production.”

This move reflects a broader market shift. Industry analysts note that businesses are rapidly moving beyond simple conversational AI and toward goal-driven, autonomous systems. The next frontier is not just having agents, but orchestrating them effectively to turn fragmented processes into intelligent, adaptive workflows that span an entire organization.

Taming Complexity at Hyperscale

Processing 45 billion tokens every day is a monumental task. This torrent of data fuels everything from network monitoring and fraud detection to customer service interactions. For AT&T, managing this volume efficiently while improving accuracy and controlling costs is a primary strategic driver. The company has increasingly adopted fine-tuned SLMs to optimize latency and has seen cost reductions of up to 90% in certain applications.

The H2O AI Super Agent platform is designed to build on this success. By unifying enterprise-grade Retrieval-Augmented Generation (RAG) pipelines, predictive systems, and language models under one architectural roof, AT&T aims to achieve new levels of efficiency and capability. The platform will be deployed across several critical business areas:

• Fraud Prevention: By combining predictive intelligence with agentic reasoning, the system can identify subtle anomalies and complex financial crime patterns in real-time, reducing risk far more effectively than static rule-based systems.
• Customer and Field Operations: Fine-tuned language models will power more intelligent customer support and field operations, improving resolution quality, response speed, and overall operational efficiency.
• Enterprise Research: Autonomous agents will be able to “talk” to documents, databases, and other data sources, enabling deep research, analysis, and synthesis that can inform high-stakes business decisions.
• Data Analysis: The platform will provide advanced analysis and visualization of structured data, including the vast network and tower datasets that are core to AT&T's business.

This enterprise-wide deployment demonstrates a commitment to embedding autonomous intelligence directly into the operational fabric of the company, moving AI from a supporting role to a central driver of business value.

Building Trust with Sovereign AI

For a company in a highly regulated industry like telecommunications, innovation cannot come at the expense of security and trust. This is where H2O.ai's 'Sovereign AI' approach becomes a critical enabler. This framework is designed to give enterprises full control over their data, models, and AI execution, addressing paramount concerns around data privacy, security, and regulatory compliance.

The H2O AI Super Agent platform supports sovereign deployments, including air-gapped and on-premises configurations. This ensures that AT&T's sensitive customer and operational data can remain entirely within its own secure infrastructure, isolated from external networks and shielded from unauthorized access. This is not just a feature; it is a foundational requirement for operating in data-sensitive environments governed by regulations like GDPR and CCPA.

Furthermore, the system is built from the ground up with embedded risk controls, transparent audit trails, and robust governance features. Every decision and action taken by the AI agents is traceable, providing the accountability necessary for compliance and internal oversight. This 'trust factor' is what allows enterprises like AT&T to confidently deploy powerful autonomous systems for mission-critical tasks.

A Partnership Forged in Production

AT&T’s adoption of the Super Agent is the latest chapter in a long-standing collaboration with H2O.ai, focused on operationalizing AI in demanding, large-scale production environments. The two companies previously co-developed the H2O AI Feature Store, a platform designed to help teams reuse, govern, and operationalize machine learning features across an enterprise—addressing the resource-intensive challenge of feature engineering at scale.

This history of tackling difficult, real-world AI problems has built a strong foundation for this ambitious new phase. The partnership extends beyond commercial applications, including joint AI-for-Good initiatives such as flood prediction research with Argonne National Laboratory, proving that AI-driven models can accelerate critical scientific discovery.

“AT&T stands out as one of the most advanced enterprise AI organizations in the world – we are excited to be powering thousands of users and agents on Ask AT&T,” said Sri Ambati, Founder and CEO at H2O.ai. “H2O AI Super Agents are sophisticated coding agents with accurate tool calling, orchestrating deep research, reasoning, predictive intelligence, and purpose-built small language models. They are designed for traceable, high-fidelity execution across long-running business tasks on both private and public data.”

As AT&T integrates these powerful new capabilities, the industry will be watching closely. This deployment is more than just a technology upgrade; it is a blueprint for how the world's largest companies can harness the power of autonomous AI to compete and thrive in the 21st century.