AT&T keeps a person checking nearly 500 AI workflows until each hits mid to high 90% accuracy. SAP delivery leads sign go-lives.
AT&T has nearly 500 AI agent workflows in production.
It plans to triple that next year, SiliconANGLE reported on 14 November. Luke Corbin of AT&T Business spoke at an AWS event in Boston.
Every production workflow starts under human oversight. An agent runs on its own only after its accuracy reaches the mid to high 90% range within an accepted risk profile. Workflows that touch customers pass security, privacy and compliance reviews first.
Corbin said production AI investments are achieving fivefold returns within a year. The report gives no spend figure or baseline.
A December 2025 Harvard Business Review Analytic Services study of 603 leaders found just 6% of companies fully trust AI agents to run core business processes on their own. Only 12% said their risk and governance controls were adequate.
For SAP delivery leads, the AT&T rule turns "the agent works" into a test result someone has to sign. On an ERP programme that is usually the person who signs the cutover and runs the support weeks after it. Our reading: the go-live plan now needs an accuracy bar and a method for measuring it. It also needs to say how long a person keeps checking.
Sources
Our file on AT&T
- 30 Sept
AT&T moved 40% of its AI work to cheaper open models. It wants 70%.
- 12 Sept
An AT&T store worker swapped customers' SIMs for cash. One victim lost $99,528.
For your next steering committee
- 1Ask which design decisions in your programme would be expensive to reverse, and when each was last reviewed.
- 2Ask which expected benefits change if the implementation approach changes.
- 3Ask who owns a decision to change course, and where it would be recorded.
Issue 01 of the ERP LEADERS brief puts a timeline, the three-number review and six questions on one page. Read issue 01
Welcome back. · Issue 01



