OpenAI pushes enterprise plumbing

OpenAI is quietly changing what it sells. The new release is not a flashier chatbot voice or a benchmark chart; it is a way to make custom ChatGPT bots call outside tools on the web using GPT-4o and GPT-4.1 for paid users on Plus, Pro, and Team plans. (help.openai.com) A Custom Action is basically a buttonless software bridge. Instead of copying text from ChatGPT into another app, a custom bot can send a request straight to an outside service through an application programming interface, which is the standard pipe software uses to talk to other software. (developers.openai.com) OpenAI already had these actions, but the new release notes say they now work with GPT-4o and GPT-4.1 on the web for Plus, Pro, and Team users, while Enterprise and Education access is listed as “coming soon.” That shifts the feature from a niche builder tool toward something ordinary paid users can actually use in a browser tab. (help.openai.com) The company has been building the rest of that plumbing in parallel. OpenAI’s apps page says ChatGPT can now pull context from business tools and data so users can create, analyze, and take action in one place, which is a very different pitch from “ask a smart model a question.” (openai.com) The naming tells the same story. OpenAI’s help center says “connectors” were renamed to “apps” on December 17, 2025, and those apps are described as a way to search, reference, and work faster without leaving the conversation. (help.openai.com) That matters because a model is only half the product in a workplace. A company buying ChatGPT for sales, support, or engineering usually needs three extra pieces: access controls, links to internal tools, and a bill that finance can predict. (help.openai.com) OpenAI’s new Codex rate card tackles the billing piece directly. The help article says that as of April 2, 2026, Codex pricing moved from per-message charges to token-based pricing for new and existing ChatGPT Business customers and for new ChatGPT Enterprise plans. (help.openai.com) A token is the meter on the side of the machine. Instead of charging for one “message,” OpenAI now charges based on how much text and tool use the system actually processes, which is closer to how cloud computing and the OpenAI application programming interface already work. (help.openai.com) OpenAI says older Plus, Pro, and existing Enterprise and Education customers stay on the legacy Codex rate card for now and will be migrated in the coming weeks. That means two companies using the same product can briefly be on different billing logic depending on when and how they bought it. (help.openai.com) The release notes for ChatGPT Business add one more clue. OpenAI says it updated the Codex rate card to align with token-based usage pricing, while regular ChatGPT subscription rates stayed unchanged, which separates the seat fee from the heavy-usage work happening underneath. (help.openai.com) Put together, this looks less like a company selling one magic model and more like a company selling a stack. The stack now includes seats, actions, apps, workspace controls, and usage credits, which is the kind of product shape large companies expect from software they plan to wire into daily operations. (help.openai.com)