Enterprise AI is going 'agentic'
AI spending is moving from experiments into production workflows where models act on tasks, not just answer prompts — OpenAI says enterprise now accounts for over 40% of its revenue. (decrypt.co) Companies building models are competing on enterprise controls and pricing, and even talent is shifting toward product‑level AI work — a former Workday CTO moved into a technical staff role at Anthropic to bring enterprise data experience into model development. (siliconangle.com) (thenextweb.com)
A year ago, a lot of companies were paying for artificial intelligence that worked like a very smart search box. This week, OpenAI said enterprise customers now make up more than 40% of its revenue, and it tied that jump to “agentic workflows,” where the software carries out multi-step work instead of stopping at an answer. (openai.com) That shift changes what buyers are purchasing. A chatbot writes a draft when you ask; an agent opens files, pulls data from internal systems, updates documents, and keeps going through the task list on its own. (anthropic.com) OpenAI’s own pitch now sounds less like office software and more like operating infrastructure. In its April 8 note, the company said customers want one “underlying intelligence layer” connected to internal systems, external data, and company permissions, not a pile of separate artificial intelligence tools. (openai.com) The reason big companies care about permissions is simple: an agent can do more damage than a chatbot if it has the wrong access. OpenAI’s business and enterprise plans now emphasize custom retention policies, encryption at rest and in transit, and no training on business data by default. (openai.com) Anthropic is chasing the same buyers with the same message, but from the controls side. SiliconANGLE reported on April 9 that Anthropic added organization-wide controls for Claude Cowork, including role-based access controls, group spend limits, and monitoring hooks for security pipelines. (siliconangle.com) Those controls exist because agent software is moving out of engineering teams. Anthropic says Claude Cowork was built after non-technical teams such as Marketing and Data kept bypassing chat and using more autonomous tools for research synthesis, document preparation, and file management. (anthropic.com) Pricing is changing with the product. OpenAI’s help center says ChatGPT Business and ChatGPT Enterprise added a new Codex-only seat type on April 2, 2026, so a company can pay for employees who mainly need coding agents without buying the full package for everyone. (help.openai.com) OpenAI is also widening the funnel for heavy users. SiliconANGLE reported that the company introduced a $100-per-month Pro plan for Codex users, half the old top-end price, and positioned it against rival coding products that start at $200 per month. (siliconangle.com) The talent market is moving in the same direction as the product market. The Next Web reported on April 10 that Peter Bailis, who had been Workday’s chief technology officer since May 2025, left for Anthropic as a member of technical staff to work on reinforcement learning. (thenextweb.com) That is a strange title change only if you think the center of gravity is still the old software stack. If the valuable work is shifting into the model layer itself, then an executive with enterprise data experience can have more leverage helping train and shape the system than running a traditional software org chart. (thenextweb.com)
