Antimetal's ops lead admits the team switched off Claude internally

Antimetal’s operations lead says the company has moved its internal AI stack from Claude to ChatGPT and Codex. That’s a useful reminder for any business owner building AI into day-to-day work. The tool your team starts with doesn’t have to be the tool you keep. What matters is whether it helps people do real work reliably, fits the way your team operates, and gives you enough control over cost and output.

The more interesting part may be how Antimetal is testing its own pricing. The company already publicly prices around success, taking a share of cloud savings delivered. It’s now live-testing usage-based pricing against outcomes-based pricing. Those are two very different promises. Usage pricing is straightforward, you pay for consumption. Outcomes pricing puts more of the burden on the provider to prove value. For customers, that can feel lower risk, but it also requires clear measurement and agreement on what counts as a result.

This is where AI business models are heading for many operators. Don’t just ask what an AI tool costs per seat or per token. Ask what business result it is tied to, who measures that result, and whether the supplier has any real accountability if it doesn’t show up. This is the kind of thing we build into an AI command centre, connecting AI activity to the operational measures that actually matter.