OpenAI preps one ChatGPT for chat and work as usage bills bite

OpenAI wants ChatGPT to stop being just a question-and-answer box. The company is planning tests that blend its Work business agent into the same interface as regular chat, so people can move from asking something to finishing a project without hopping between screens. Testing starts with a small group of users, according to an email seen by BleepingComputer.

Today the two jobs are split. Chat handles quick questions and brainstorming, while Work takes on research, spreadsheets, reports, presentations and published Sites. A merger would keep file uploads and connected plugins inside one conversation, with users still able to watch and steer the agent's output. Codex, for its part, stays a standalone platform.

Anthropic got there first. Its September 16 announcement was blunt: Cowork and chat are merging into one Claude. That unified experience began rolling out to Pro, Max and Ultra subscribers, Enterprise joined on October 23, and Team and Free plans are still waiting their turn. Anthropic also introduced Claude Docs and Slides for editing documents and presentations.

Every extra task runs on a meter. Work and Codex already share an allowance structure, and credits can be bought once it runs out. On the Pro $500 plan, Ultrafast eats included usage at eight times the standard rate, then draws credits at six times. December brings another change: Plus, Pro and Business subscribers will share a token-based allowance across ChatGPT and Codex, with instant chat exempt, and users will be able to check remaining allowance and reset times under Settings > Usage.

The direction of travel has been generous until now. OpenAI said on October 29 that it doubled Work and Codex allowances for Pro $100 and $500 subscribers, and Pro $200 accounts also got higher limits. Gartner expects spending on AI agents and assistants to reach $65.5 billion in 2027, up from $29.2 billion in 2026, while total AI spending could hit $2.67 trillion this year, a 49.5% jump.

Adoption, though, is lopsided: an IDC study of 1,900 organizations found only 3.1% at the highest AI-maturity stage, with 61.3% in the two earliest. Predictable pricing could decide how quickly that gap closes. Cryptopolitan reported Codex users complaining their allowances drained unusually fast, and OpenAI executive Thibault Sottiaux disputed allegations of undisclosed cuts. Kingy AI initially rated the claims 'supported but unproven,' later shifting to 'partly confirmed; blanket cut unproven' after acknowledging cache-efficiency problems.