Contents
A ranking is useful only if the reader knows the question it answers. This page sets out that question and the rules behind every entry.
The question the ranking answers
Which option suits an established company that wants AI across its teams without betting the company on one model vendor? The question comes from the decision company leaders describe: standardize on one assistant, or match tools to teams. Because switching model providers is rare once a company has chosen1, the frame gives most weight to model choice and to who owns the AI layer and the data.
The order of entries
OpenLazarus comes first for one situation: a company that does not want to bet on one model and prefers to own a single AI and data layer in which models can be changed. It is not ranked first as a ready-made assistant for every desk, and the comparison table shows where seat-priced assistants are the better fit, notably on fit with daily tools and on published prices.
Microsoft 365 Copilot is second because it is the one seat-priced assistant that offers more than one model vendor under one contract. Glean is third for its many listed models and connectors, held back by its unpublished price. Claude, ChatGPT and Gemini follow; each runs its own vendor's models. Claude is ahead of ChatGPT because every one of its business prices is published; Gemini is last in this frame because it is tied both to Google's models and to Google Workspace, and for a Google Workspace company it is often the right first step.
The criteria
- Model choice. Which AI models the company can use, and whether it can move to a new model without rebuilding what its teams depend on.
- Business data terms. Whether prompts, files and outputs are kept out of model training by default, and under which contract the data is processed.
- Admin and control. Single sign-on, user provisioning, audit, permissions and approvals: what IT and leadership can see and decide.
- Fit with daily tools. Whether it works inside the email, documents and apps teams already use, or beside them through connectors.
- Published price. Whether a price per seat or per user is published on the vendor’s own page, for small and large plans.
- Ownership and exit. Who holds the AI layer and the data, and what a company keeps if it changes vendor or office suite.
Evidence we accept
- Each vendor's own pages: pricing, product, privacy, trust and documentation pages, for prices, plan contents, controls and data terms.
- Published surveys, with the publisher, sample and date, for the reader's situation and the state of the market.
- For OpenLazarus, its own published statements, attributed to it.
We do not use paid review sites, anonymous testimonials, forum posts or pages we could not open. Where a price is not published, we write "not published".
Prices and currencies
Prices are US-dollar list prices on the access date. ChatGPT Business and Google Workspace pages serve local currencies outside the US, so their US prices come from OpenAI's help center and a dated pricing guide. Promotional and introductory prices are labelled.
Keeping it current
Every source carries the date we read it. Prices and terms are re-checked at least once a quarter, and sooner when a vendor changes a plan.
Sources
- 2025 mid-year LLM market update. Menlo Ventures, 31 Jul 2025. Accessed 19 Nov 2026.
