AT&T AI Routing Cuts Inference Costs; Tools4AI Brings On‑Premise Agents to Insurance

AT&T Inc. unveiled an intelligent AI routing framework that dynamically selects among multiple open‑source language models to optimise cost, latency and response quality. The system, built around an AI gateway, can shift requests to smaller models for routine tasks and larger models for complex work, reporting up to a 90% reduction in inference expenses. Supported models include Google’s Gemma 4, OHI‑4, OSS‑120B and a telecom‑specific model called OTel 2.0, trained on extensive telecom datasets under the ModaaS program.

Tools4AI, a pure‑Java agentic AI framework, paired with the Ollama platform for local model execution (e.g., Llama 3.1, Phi‑4), enables enterprises in regulated industries such as insurance to run AI agents on‑premise. A tutorial demonstrates a claims‑triage agent that ingests free‑text incident reports, routes them to appropriate business actions, extracts structured data, and enforces human approval and audit trails, ensuring that sensitive data never leaves the organization.