Enterprise AI Offerings

The category is defined by the purchase shape: AI sold as business software. The 5 entries are not inference endpoints with rate tables — they are seat plans and contracts: GitHub Copilot at its Business and Enterprise tiers, Microsoft 365 Copilot as an add-on to the Microsoft 365 suite, Google's Gemini at its Business and Enterprise Workspace tiers, DataRobot's agent-workforce platform, and one further agent-platform vendor whose details live on the entry page. That is why the per-1M price columns read empty for most of these rows and the free-credit column carries only near-zero or unquantified values: the billing unit is the seat or the contract, not the token.

They are tracked in the same directory as the self-serve APIs for one reason: the model landscape. The model families these products route — the OpenAI and Anthropic models inside Copilot, Google's own models inside Gemini — are the same families tracked entry by entry in the paid-api category. An enterprise offering changes how you buy and govern access to those models; it does not, by itself, add new models. The directory records both sides so a reader can see the same model as a per-token API and as a seat-based business product, with the different paper attached to each.

Four shapes cover the category. Seat plans with pooled credits: GitHub Copilot is the cleanest documented example — Free at $0 with 2,000 completions per month on smaller models, Pro at $10 per user per month, Pro+ at $39, Max at $100, then Business at $19 and Enterprise at $39 per user per month, where the plan price includes pooled API credits (1,900 per user per month at Business, 2,000 at Enterprise, shareable within the organization) on top of the agent features. Suite add-ons: Microsoft 365 Copilot and Gemini Business attach AI to an existing workspace — the Microsoft pricing page documents the plan family (individual, business, enterprise), and the Google Workspace pricing page documents Gemini availability across the Business SKUs: limited access on Starter, full in-app Gemini on Standard, Plus and Enterprise, with Starter/Standard/Plus capped at 300 users and Enterprise unbounded. Agent platforms: DataRobot's positioning is an agent workforce platform — the pricing page documents trial and demo paths and on-prem, hybrid and cross-cloud deployment, with licensing that is quote-based rather than self-serve. The fifth entry follows the platform shape as well, with its specifics on the entry page.

The shape determines what you can compare publicly. Seat plans publish prices (the Copilot tiers above); suite add-ons publish plan availability but route real pricing through the suite's sales motion; quote-based platforms publish almost nothing. The directory reflects that gradient honestly: where a number is published it is tracked, where it is not, the entry says the purchase path instead of inventing a figure.

The useful read of this category is as a map of where the models you already know are also available by contract. Copilot's own plans page documents which models sit in which tier — the free tier runs smaller models (Haiku 4.5 and GPT-5 mini at verification), the paid tiers add the frontier models, and the business tiers add governance (pools, SSO, billing controls) on top of the same model access. Gemini Business is the Workspace distribution of Google's models — the model family is the one tracked in the official-free-tier and paid-api categories; the SKU determines the seat count and the in-app surface. M365 Copilot is the Office-surface distribution of the same OpenAI model families that the paid-api category tracks at token rates. The directory value is the side-by-side: the token rate on one page, the seat price on this one, for a model that is materially the same.

The "different paper" is not just legal texture. Pooled credits change who can spend what (an organization pool, not a personal meter); suite deployment changes where the AI lives (inside the inbox, the doc, the repo — not in your client code); governance changes what you can audit (pools, seat controls, the vendor's compliance artifacts). None of those facts is in a per-1M price column, which is why this category is a procurement page rather than a price page.

Can verify, from published pages: plan existence and published seat prices (the Copilot tiers, the Workspace SKU limits — including the 300-user cap on Starter/Standard/Plus that the Google pricing page states), trial and demo paths (DataRobot's pricing page links both), the tracked web surface's live status via the probe, and the model roster where the vendor publishes it (Copilot's plans page does). Cannot verify, and does not claim to: SLAs, data-processing agreements, compliance certifications, procurement terms, or the actual quote a vendor will send — those sit behind the sales motion, and the directory's rule is to track what is published and mark the rest as vendor territory.

The route is right when the constraint is organizational rather than technical: a team needs governed access (pools, SSO, centralized billing), the AI must live inside the suite the team already runs (Office, Workspace, the repo), or the deployment has to be on-prem or hybrid (the documented DataRobot shape). It is the wrong route for a product team integrating a model into its own software — that is the paid-api category, where the per-token key is the product. The common end state is the hybrid: self-serve keys for the application layer, enterprise seats for the people layer, both visible in this directory so the total spend and the model parity can be checked against one source. Every entry here has that self-serve sibling, and the entry page links it — the sibling's published rate is the public reference for what the contract is actually wrapping.

Five checks. One, purchase shape: seat plan with published price (Copilot), suite add-on with sales motion (M365 Copilot, Gemini Business), or quote-based platform (DataRobot) — each answers a different buyer question, and the entry says which one applies. Two, credit shape: are there pooled API credits on top of the seat (Copilot Business 1,900 and Enterprise 2,000 per user per month, verified), and can your workload consume them through the product surface? Three, deployment: where the AI must live — the suite, the repo, or on-prem/hybrid (the documented DataRobot options) — because that constraint usually decides before price does. Four, model parity: which model families the offering exposes, checked against the same families' paid-api entries — an enterprise contract that hides the models your app actually needs is not a saving, it is a lock. Five, trial path: a $0 or trial tier (Copilot Free, DataRobot's trial link) exists on most rows here and is the cheap way to test the surface before the procurement conversation.

What this page does not decide: the actual quote, the compliance fit, or whether your organization's suite commitment already includes the AI — those are vendor conversations. It decides the smaller facts: which offerings publish seat prices, what the published credit pools are, where the trial paths are, and what the tracked surface is doing right now.