ChatGPT Work — New Enterprise Tier
OpenAI just made its business pricing
a three-tier stack.
OpenAI's business lineup, previously just Business and Enterprise, now has a top tier: ChatGPT Work. According to OpenAI's announcement, Work bundles auditing, SSO, department-level billing, and reserved capacity. As the IPO push accelerates, a new line quietly appeared on the approval sheet — we read the structural change through the eyes of procurement, IT, and PdMs.
What Changed
"One more step above Business" — what it really means
Not a new plan, a new purchasing lane.
Until now, OpenAI's business lineup was a two-tier stack: Business, seat-based and easy for teams to swipe onto a credit card, and Enterprise, the whole-company contract with audit logs and SSO. The newly launched ChatGPT Work sits above both. Per OpenAI's own description, Work bundles four things: department-split invoicing, reserved inference capacity, an audit-DLP-SSO one-package, and a rollout SLA — effectively an "Enterprise full suite" tier.
The point isn't a new plan label; it's the separation of purchasing paths. Business gets in through team credit cards, Enterprise through HQ contracts, and Work through "department-level budgets with department-level approval." Between the grassroots Business path and the top-down Enterprise path, OpenAI inserted a department-scoped governance layer. That's the more accurate framing.
The Numbers
The three tiers, side by side
| Business / Enterprise (existing) | ChatGPT Work (new) |
|---|---|
| Company-wide contract assumed | Budget & invoicing split per department |
| Shared inference pool | Reserved Capacity per department |
| Audit & SSO are add-ons | Audit logs, SSO, DLP bundled by default |
| Rollout is IT's job | Department-led rollout with an SLA |
Why It Matters
Underneath the IPO run, they're layering on revenue thickness
This is not a feature drop. For the first time, OpenAI has explicitly said "Business, Enterprise, and Work can coexist inside the same company." The old sales pitch was linear — "start small, grow into a company-wide contract." Work opens a second upsell lane: "layer deeper, per-department contracts on top of Enterprise."
In pre-IPO disclosures, gross margin and net revenue retention (expansion revenue from existing customers) are the metrics investors interrogate. ChatGPT Work is best read as a setup to book "layered spend" from existing Enterprise customers. The fact that the pricing page lists Work alongside Enterprise (not replacing it) is corroborating evidence: this is stacking, not substitution.
Who It Hits
Whose approval sheet does this actually touch?
The impact is opposite for IT, procurement, PdMs, and end users.
IT & procurement
One more approval tier to model. You keep running Business and Enterprise while also wiring department-originated Work requests into your approval flow and SSO governance — that's real integration work.
Department heads & PdMs
Biggest win: you can run AI adoption on your own budget without waiting on IT. But you also inherit the governance load — DLP and audit setup become the department's problem.
End users on Plus / Free
Zero impact. Same UI, same models. The change is confined to the invoice hierarchy and admin panel — nothing that affects the person actually using ChatGPT.
Work isn't selling features.
It's selling a purchasing lane.
What's Next
What comes next / what to do about it
Take stock
Audit your Business/Enterprise contracts and flag departments that genuinely need split invoicing — legal, R&D, overseas offices. That's where Work should land, and only there.
Redesign the controls
Work bundles SSO, audit, and DLP — but if that overlaps with your existing Enterprise controls, policies will conflict. Decide where logs flow and how identity is scoped before switching on Work.
"Extend Enterprise, or stack Work on top?"
Renewal-cycle Enterprise customers will get the Work upsell pitch from OpenAI. Extend vs. stack should be decided by how independent department budgets really are, not by feature parity.
The Counterpoint
The other side: this structure has a cost
Don't cheer this uncritically. More plan tiers means license optimization gets harder. When Business, Enterprise, and Work run in parallel inside the same company, users who work across departments frequently end up double- or triple-counted. That failure mode has already been documented on other multi-tier SaaS — treating it as a hypothetical is naive.
Reserved Capacity sounds attractive, but if actual usage falls short, it just becomes idle spend. If you're using Reserved as your justification for approval, measure your real rate-limit headroom for one or two months first, then upgrade. Procurement should share the framing early: a higher tier isn't "more features" — it's "more operational responsibility."