AI vendors shift to enterprise selling

OpenAI and Anthropic are starting to sell artificial intelligence less like a magic trick and more like payroll software. In the first week of April 2026, OpenAI added flexible seat types for ChatGPT Business and ChatGPT Enterprise, while Anthropic kept pushing Claude with admin controls and enterprise plans built for large organizations. (openai.com) (anthropic.com) That change shows up first in pricing. OpenAI said on April 2, 2026 that Business and Enterprise workspaces can now mix a standard ChatGPT seat with a new Codex-only seat, so a company can buy cheaper access for employees who only need the coding agent instead of paying for the full bundle. (openai.com) Big companies do not buy software the way consumers do. A procurement team usually asks who can log in, who can see logs, how data is retained, and whether costs stay predictable across 5,000 employees, which is why vendors are now talking about identity controls and billing structures instead of just benchmark scores. (openai.com) (anthropic.com) OpenAI has been filling in that checklist piece by piece. Its enterprise privacy page says business customers own and control their data, model training on that data is off by default, retention can be controlled on ChatGPT Enterprise, and security includes Security and Compliance 2, or SOC 2, plus encryption at rest and in transit. (openai.com) It also added the kind of record-keeping compliance teams expect. OpenAI said its Compliance Logs Platform includes immutable JavaScript Object Notation Lines log files with minutes-level latency, and the December 11, 2025 update added Admin Audit, User Authentication, and Codex Usage logs. (openai.com) Anthropic has been making the same pitch in its own language. Its Claude Enterprise materials list single sign-on, domain capture, audit logs, System for Cross-domain Identity Management provisioning, and role-based permissions, which are the controls an information technology department uses to keep one employee’s chatbot from turning into an unmanaged company system. (anthropic.com 1) (anthropic.com 2) Anthropic added another layer in August 2025 when it rolled out a Compliance Application Programming Interface for Team and Enterprise customers. The company said that interface lets organizations export audit logs to external monitoring tools, which is the sort of feature security teams ask for before approving wide deployment. (anthropic.com) This is happening because the money is shifting toward business accounts. Axios reported on March 18, 2026 that Anthropic was capturing more than 73 percent of spending among companies buying artificial intelligence tools for the first time, using Ramp customer data as the measure. (axios.com) The rivalry is now sharp enough that it is spilling into investor messaging. CNBC reported on April 9, 2026 that OpenAI sent shareholders a memo attacking Anthropic’s scale and compute position, which suggests the fight is no longer just about who has the flashiest demo but about who can lock down the biggest contracts. (cnbc.com) The headline this week is not that either company built a suddenly smarter model. It is that both are packaging artificial intelligence so a chief information officer can approve it, a security team can monitor it, and a finance department can budget it without guessing. (siliconangle.com)