344
Productivity & Workflow355
Automation & Workflow225
Software Development251
Marketing & Growth192
AI Infrastructure & MLOps175
Writing & Content Creation203
Data & Analytics142
Photography & Imaging156
Design & Creative170
Customer Support132
Sales & Outreach125
Voice & Speech135
Education & Learning131
Operations & Admin87
OpenAI is expanding ChatGPT Work, a tool that can take actions in apps like email and calendars, aiming to make AI “agents” useful beyond coding.
In short: OpenAI is trying to make “AI agents” that can do multi-step office tasks, not just answer questions or write code.
OpenAI is promoting a product called ChatGPT Work, which it released last month for $20 per month. The idea is to let people use AI agents, meaning an AI tool that can take actions across your work apps, like an assistant who can do tasks for you, not just give advice.
In the TechCrunch report, an OpenAI engineer said he lets the desktop app access and control tools like his inbox, Slack, and other work apps. That access is the point, because the AI needs permission to read information and click buttons, similar to giving someone a key so they can tidy your house while you are out.
OpenAI says getting regular office workers to use agents is harder than getting software developers to use them. An OpenAI-backed study found that 98% of OpenAI employees used its Codex agent tool in June, but only 17% of organizational customers and under 1% of individual customers did. That gap shows why OpenAI is trying to make the product simpler, with more buttons and clearer steps.
The report also describes real uses and real limits. ChatGPT Work could copy a preschool calendar from email into Google Calendar, but setting permissions was confusing, and some features were missing, like creating new calendars.
For most people, the big question is trust. These tools work best when you give them access to private accounts, and mistakes could expose sensitive information. Another issue is cost, since longer tasks use more “tokens” (units of text the AI reads and writes), which can be expensive behind the scenes.
Source: TechCrunch AI