8 billion tokens a day forced AT&T to rethink AI orchestration — and cut costs by 90%
Summary
AT&T's innovative Ask AT&T Workflows, developed by Chief Data Officer Andy Markus, utilizes a multi-agent stack to enhance efficiency, achieving up to 90% cost savings. This AI-driven approach empowers over 100,000 employees, streamlining task automation and software development.
Key Insights
What are tokens in the context of AT&T's AI platform?
Tokens are small units of text, roughly equivalent to words, that AI models process. AT&T's Ask AT&T platform handles massive volumes, with reports of 5-9 billion tokens daily, driving high-scale AI operations.
How did AT&T achieve 90% cost savings in their AI operations?
AT&T achieved up to 90% cost reduction by fine-tuning small language models (SLMs) with proprietary knowledge, improving efficiency over larger models while maintaining accuracy at scale.
Sources:
[1]