OpenAI's Codex pay-as-you-go seats lower the enterprise drawbridge

In enterprise software, the biggest barrier to adoption isn't always model capability; sometimes it's just the billing structure. In its Codex flexible pricing announcement, OpenAI outlines how ChatGPT Business and Enterprise teams can now add Codex-only seats with pay-as-you-go pricing, completely sidestepping fixed seat fees.

While the source page's audit was blocked during our review pass, the mechanics from the original announcement are straightforward: usage is billed by token consumption rather than a flat per-user tax, and these Codex-only seats operate without rate limits.

This is less about generosity and more about reducing activation energy. Coding-agent rollouts often stall before anyone even tests the model because managers need a small group to run a narrow workflow without triggering a massive procurement fight. By offering $100 in credits for new Codex-only team members and lowering standard Business annual pricing, OpenAI is making the wedge explicit.

Pay-as-you-go pricing also forces a healthy kind of discipline. If teams are running Codex constantly and can't connect the token spend to accepted diffs and shipped code, the invoice becomes a very blunt product review. It lets organizations treat the agent like a teammate on probation: define candidate tasks, set success criteria, track the spend, and expand only where the AI actually reduces engineering friction.