Superpower Daily
NVIDIA Adds BlueField-4 Scale-In for 800 Gb/s AI-Factory Infras...Nvidia Puts Groq 3 LPX Into Production for Faster AI-Agent Resp...OpenAI Pauses Frontier Training After Sandbox Breach, Warns of...Cloudflare Open-Sources Cloudflare OS, Putting AI-Built Work Ap...AWS Backs ARD Open Spec to Federate AI-Agent Discovery Across E...UC San Diego Will Test $8.48M System to Make 2 MW of AI Compute...NVIDIA Adds BlueField-4 Scale-In for 800 Gb/s AI-Factory Infras...Nvidia Puts Groq 3 LPX Into Production for Faster AI-Agent Resp...OpenAI Pauses Frontier Training After Sandbox Breach, Warns of...Cloudflare Open-Sources Cloudflare OS, Putting AI-Built Work Ap...AWS Backs ARD Open Spec to Federate AI-Agent Discovery Across E...UC San Diego Will Test $8.48M System to Make 2 MW of AI Compute...
Productspublished

OpenAI’s $20 ChatGPT Work Brings Codex Agents to Office Apps, With Access Still a Barrier

The product’s promise is a ChatGPT that acts across workplace systems rather than merely answering prompts. Its immediate test is whether non-engineers will grant access and navigate controls built for more autonomous software.

By 3 min read
OpenAI’s $20 ChatGPT Work Brings Codex Agents to Office Apps, With Access Still a Barrier
Sign in to save and react

Listen to this story

The audio brief

About 1:39
0:001:39
Read transcript
ChatGPT Work is already showing the central problem with workplace AI agents: giving them access can be harder than asking them to do the job. In testing, OpenAI’s twenty-dollar-a-month product could create an event in Google Calendar, but it could not create a new calendar. Some cloud-drive connections reportedly demanded complete access even when a user only needed read-only access, and certain controls were available only on the web. ChatGPT Work is an adaptation of Codex, OpenAI’s coding agent, for non-engineers. It can connect to email, Slack, Notion, Figma, Salesforce, Google Calendar, and web browsers, so it can work across several systems instead of just answering a prompt. But the adoption data behind that ambition is striking. An OpenAI-backed study found Codex usage at 98 percent among OpenAI employees, compared with 17 percent among organizational subscribers and less than 1 percent among individual subscribers. OpenAI has not released separate adoption figures for ChatGPT Work. The software layer controlling an agent’s permissions, tools, context, and results is called a harness, and OpenAI says its setup is not yet intuitive for new users. The company has made Codex more interactive after learning that its earlier design moved faster than the model and harness could reliably support. The harder test may be business work, where GDPval covers 44 occupations, but presentations and strategy are more difficult to evaluate than code. The key constraint is whether ordinary office users will grant enough access—and trust the controls once they do.

Story brief

3 key points

OpenAI is testing whether its coding-agent playbook can extend to everyday office work with ChatGPT Work, a $20 monthly product that operates across business apps. The market challenge is adoption and control: Codex usage reached 98% inside OpenAI but only 17% among organizational subscribers and under 1% among individuals. Early testing found confusing permissions, web-only settings, and limited calendar actions....

  1. 01

    ChatGPT Work can connect to email, Slack, Notion, Figma, Salesforce, Google Calendar, and browsers.

  2. 02

    Testing found Google Calendar event creation, but not the ability to create a new calendar.

  3. 03

    Some cloud-drive integrations reportedly demanded complete access despite read-only use cases.

ChatGPT Work asks office workers to give an AI system access to the software that holds their schedules, messages, documents, and customer records. OpenAI created the $20-per-month product by adapting Codex, its coding tool, for non-engineers who want an agent to complete multistep work across those systems.

The ambition is to move ChatGPT beyond a prompt-and-response interface. Work can connect with email, Slack, Notion, Figma, Salesforce, Google Calendar, and web browsers, giving an agent the tools and context to carry out longer tasks instead of simply suggesting how a user might do them.

The adoption gap behind the workplace pitch

OpenAI has already seen the difference between building an agent for technical users and getting one adopted more widely. An OpenAI-backed study measured Codex use at 98% among OpenAI employees in June, versus 17% among organizational subscribers and less than 1% among individual subscribers. OpenAI has not disclosed separate user counts for Work and Codex.

Codex use was far higher inside OpenAI
98%OpenAI employees using Codex

The OpenAI-backed study measured employee use in June.

17%Organizational subscribers using Codex

Organizational-subscriber use was substantially below the internal rate.

Less than 1%Individual subscribers using Codex

Individual-subscriber use was below 1% in the same study.

More access creates a harder control problem

The product depends on a harness: software around a language model that determines what information it sees, which tools it can use, and how it returns results. For agents, that layer also provides instructions and tools for longer tasks. Its design now becomes a practical constraint on Work’s usefulness.

The current limits are concrete

  • Testing described confusing cloud-drive permission setup, including read-only access errors and a prompt saying complete access was required. Some settings were available only on the web app.
  • In the article’s testing, Work could create Google Calendar events but could not create a new calendar.
  • OpenAI’s harness engineering lead said the effort setting was not yet intuitive for new users.

Coding offered a lesson, not a complete template

OpenAI has already altered Codex after Claude Code emphasized a back-and-forth workflow: present options, take another step after a user chooses, and provide continual updates. OpenAI later added more opportunities for user interaction after concluding its earlier design had moved ahead of what the model and harness could reliably handle.

That revision may be more difficult outside programming. OpenAI uses its GDPval benchmark, spanning 44 occupations and hundreds of knowledge-work tests, alongside user feedback. But presentations, sales pitches, and business strategy are harder to evaluate and trace than code, leaving the reliability of broad workplace automation unresolved.

Editorial analysis

Our Read

Our Read: The important shift is not that ChatGPT can connect to another workplace app. It is that OpenAI is trying to turn the agent harness—the controls and context around a model—into a mainstream product. The June Codex usage split shows how far that transition has to travel: internal adoption was high while individual-subscriber use was below 1%. Watch for Work-specific usage data and whether permission flows can support useful, limited access rather than pushing users toward complete access. Those are more meaningful adoption tests than the $20 entry price.

Related coverageOpenAI Lets Codex Route Smaller Tasks From Sol to Lower-Cost Luna WorkersBlitzy and XBOW Push AI Beyond Short Tasks Toward Continuous Enterprise Work

Sources

  1. techcrunch.comOpenAI is building AI agents for everything. Will everyone use them? | TechCrunch

Related coverage

Continue reading