Enterprise AI sales mature
Frontier-model vendors like Anthropic and OpenAI are lowering prices while adding enterprise-grade controls to win big customers, shifting the sales battle from pure capability to procurement readiness and predictable governance. Adoption trackers show Anthropic's paid usage climbed in March even as OpenAI maintains the largest paid footprint, suggesting procurement momentum can outpace social buzz. (siliconangle.com) (alltoc.com) (qz.com)
OpenAI and Anthropic are starting to sell artificial intelligence the way big software companies sell payroll or database tools: with admin controls, budget limits, and contracts that make procurement teams less nervous. On April 9, Anthropic added organization-wide controls to Claude Cowork, and OpenAI cut its Pro plan for Codex access to $100 a month from a higher premium tier. (siliconangle.com) That shift tells you where the fight is moving. A company with 50,000 employees does not buy a tool because a demo looked magical for 30 seconds; it buys when finance can cap spend, security can monitor activity, and managers can decide who gets which powers. (siliconangle.com) Anthropic’s new controls are aimed directly at that checklist. The company said Enterprise customers can now use role-based access controls, team spend limits, and OpenTelemetry monitoring so security teams can see tool calls and file changes inside their existing logging systems. (siliconangle.com) Claude Cowork is the product at the center of that push. Anthropic describes it as an autonomous agent that can handle multistep work on an employee’s computer, including organizing files, creating reports, and running browser tasks, which makes permission settings much more important than they are for a simple chat box. (siliconangle.com) OpenAI’s move hit the same market from the pricing side. SiliconANGLE reported that OpenAI introduced a $100-per-month Pro plan with five times the usage of its $20 tier, while Anthropic’s Claude Code and Google’s Gemini Code Assist top tiers both start at $200 per month. (siliconangle.com) The scoreboard now looks different from the social media conversation. Ramp’s March 2026 Artificial Intelligence Index said nearly one in four businesses on Ramp were paying for Anthropic a year after that figure was one in 25, while OpenAI’s adoption fell 1.5% in that March update. (ramp.com) Newer reporting suggests Anthropic kept accelerating after that. Techmeme, citing Financial Times reporting on Ramp data, said 30.6% of United States businesses paid for Anthropic’s tools in March, up from 24.4% in February, while OpenAI’s business adoption stayed near 35% month over month. (techmeme.com) OpenAI is still the bigger installed base, and it is telling investors the gap in raw infrastructure will keep it ahead. CNBC reported that OpenAI said it plans to have 30 gigawatts of compute by 2030, versus an estimated 7 to 8 gigawatts for Anthropic by the end of 2027. (cnbc.com) That argument is basically that more data-center power should make each new model stronger and each unit of output cheaper. In the same memo described by CNBC and Quartz, OpenAI said larger infrastructure and hardware improvements can lower the cost per unit of intelligence over time. (cnbc.com) (qz.com) Anthropic is betting that enterprise buyers do not wait for a distant infrastructure curve if the current product already fits how companies buy software. Its enterprise page pitches tailored terms, usage commitments, invoicing, product bundling, secure connectors, and a Health Insurance Portability and Accountability Act ready offering, which is exactly the language a procurement review expects to see. (claude.com) So the contest is no longer just which model writes the cleverest paragraph. It is which vendor can survive a security review, fit inside a department budget, plug into existing systems, and still be cheap enough that a chief financial officer signs the renewal instead of killing the pilot. (siliconangle.com)
