OpenAI’s enterprise product overhaul this week can be summarized in three slides. Here’s what they actually say, and what’s worth paying attention to.
Slide 1: The new pricing model
OpenAI moved its enterprise tier from per-seat licensing to a hybrid model: a platform access fee plus consumption-based pricing tied to API calls and agent-hours. This is a meaningful structural shift. Per-seat pricing made sense when ChatGPT was a productivity tool for knowledge workers. It doesn’t make sense when the unit of value is an AI agent completing work autonomously.
Consumption-based pricing aligns the vendor’s incentive with the customer’s: OpenAI makes more money when the agents do more work, which means OpenAI has a real incentive to make sure the agents actually work. That alignment is important in a market where enterprise buyers are getting burned by overpromised AI products.
The catch: consumption pricing is harder for enterprises to budget. CFOs don’t love it. Expect a wave of negotiated consumption caps and pre-paid credit packages as the commercial motion matures.
Slide 2: The enterprise admin console
The new admin console is the feature that matters most for IT and security buyers, and it’s genuinely differentiated. Single-tenant deployment options, fine-grained permission models (who can use which models, on what data, for what tasks), full audit trails exportable to your SIEM, and dedicated model isolation for regulated industries.
This is table stakes for regulated enterprise segments — financial services, healthcare, government. It’s also table stakes that most enterprise AI vendors don’t have. The fact that OpenAI has invested in it signals something about where they expect their revenue to come from over the next two years.
Slide 3: The SLA and the Microsoft problem
OpenAI is now offering 99.9% uptime SLAs for enterprise customers with financial penalties for non-compliance. This is the move that should make Microsoft nervous.
Microsoft has been the preferred enterprise access point for OpenAI models through Azure OpenAI Service. The enterprise SLA, sold directly by OpenAI, cuts out the Azure intermediary for customers whose primary concern is reliability guarantees. You can still use Azure if you want the full Microsoft enterprise stack integration. But if you’re primarily buying uptime assurance and want to negotiate directly, you now have a path.
The longer-term implication: OpenAI is building the enterprise relationships, contracts, and trust that let them be a direct vendor, not just a model provider to Microsoft. That’s a very different company than the one that signed the original Azure deal.