How is AT&T Using H2O AI Super Agent?
The telecommunications company processes roughly 45 billion tokens per day across its AI systems, according to its CDO.
On July 14, 2026, H2O.ai announced that AT&T is deploying its H2O AI Super Agent™ in the next phase of its agentic AI initiatives, including the Ask AT&T platform. The company said the deployment spans customer experience, fraud prevention, field operations, enterprise research, and intelligent automation.
"At AT&T, we operate AI systems at enormous scale, processing about 45 billion tokens per day across complex environments, including customer and network operations, fraud prevention, and enterprise workflows," said Andy Markus, Chief Data Officer at AT&T. "H2O AI Super Agent helps us orchestrate predictive artificial intelligence (AI) and autonomous agents in a scalable architecture built for production."
According to the company, the H2O AI Super Agent allows AT&T to orchestrate autonomous deep research agents, enterprise-grade retrieval-augmented generation (RAG) pipelines, predictive AI systems, and fine-tuned small language models (SLMs) within a single architecture.
The company said this builds on an existing relationship between AT&T and H2O.ai centered on operationalizing AI in production settings. AT&T has adopted fine-tuned SLMs to improve accuracy and latency, the company said, while reducing costs by up to 90 percent.
Beyond Single Agents to Autonomous Enterprise Systems
The H2O AI Super Agent combines large language and reasoning models with purpose-built SLMs and predictive engines. Rather than functioning as a single executor, the system acts as an orchestrator, the company said, decomposing objectives such as enterprise research and decision support into sub-tasks and deploying agents in parallel
The company said AT&T's use cases include fraud and financial crime detection, field operations intelligence, document intelligence and content generation, and analysis of structured datasets including network and tower data. The deployment also expands the Ask AT&T platform with new agent-powered capabilities, according to the company.
The system is built with embedded risk controls, audit trails, and governance for regulated environments, the company said, and supports sovereign AI deployments including air-gapped, on-premises configurations.
"AT&T stands out as one of the most advanced enterprise AI organizations in the world," said Sri Ambati, Founder and Chief Executive Officer 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."
The companies said their collaboration also includes the co-development of the H2O AI Feature Store and joint research projects, including flood prediction work with Argonne National Laboratory.
>
Key Takeaways
- Deploy H2O AI Super Agent to enhance AT&T's AI capabilities across multiple enterprise operations.
- Process 45 billion tokens daily, optimizing customer experience and fraud prevention efforts.
- Integrate predictive AI systems and autonomous agents into a unified architecture for better efficiency.
- Leverage fine-tuned small language models to improve accuracy while reducing operational costs by 90%.
- Expand use cases to include fraud detection, document intelligence, and structured data analysis.