OpenAI reveals how it actually uses ChatGPT internally

Every company that sells AI gets asked the same question eventually: are you actually using this stuff yourself? OpenAI’s Chief Commercial Officer Giancarlo ‘GC’ Lionetti just kicked off a new series called ‘OpenAI on OpenAI’ that answers exactly that, pulling back the curtain on five internal tools the company has built on its own APIs.

It’s a smart move, and not just for marketing. Lionetti frames AI adoption as a practice, not a product deployment, arguing that the real work is in encoding institutional expertise and distributing it across teams. That’s a far cry from the ‘just add API key’ fantasy of 2023. The tension he describes — deployments outpacing organizational readiness — is the same one every enterprise wrestling with this technology faces.

The tools themselves are practical rather than flashy. GTM Assistant lives in Slack and centralizes account context for sales teams. DocuGPT turns contracts into structured searchable data for finance. Research Assistant mines millions of support tickets for conversational insights. A Support Agent model uses continuous evaluations and dynamic knowledge loops to turn every interaction into training data, repositioning human reps as system builders rather than ticket handlers. And an Inbound Sales Assistant personalizes responses to every lead, handling product and compliance questions instantly before routing qualified prospects to humans with full context.

Lionetti emphasizes speed: these teams study workflows, define what good looks like, and deliver changes in weeks, not quarters. The pattern is consistent — identify high-leverage systems, test in live deployments, and build the same muscles customers need to develop. The company is releasing technical resources at DevDay on October 6 for those who want to adapt the approach.