OpenAI published two pieces this week on how businesses are putting AI to work. The first, a research overview, argues that enterprises are shifting from using AI for assistance to letting it execute tasks. It identifies a pattern where leading companies — termed "frontier firms" — are pulling ahead of others by embedding agentic AI, which includes tools like ChatGPT and Codex, into core workflows rather than treating them as add-ons.
The second piece is a case study of RingCentral, a communications company. RingCentral reports using ChatGPT Work and Codex to accelerate AI product development and to centralize operational intelligence across engineering and operations. That matches the research's emphasis on execution: the company isn't just asking questions, it's using AI to build and run software.
Where the two sources differ is in scope. The research generalizes across many enterprises, while the case study zooms into one implementation. Together they tell a consistent story: agentic AI is moving from pilots to production, but adoption is uneven — some firms are pulling far ahead of the pack.