OpenAI is moving from answers to actions
OpenAI’s latest enterprise push centers on a simple but meaningful shift: instead of using AI mainly to answer prompts, the company wants it to carry out multi-step work. With the launch of ChatGPT Work, OpenAI is positioning the product as an agentic platform that can coordinate tools, operate across applications and files, and handle long-running tasks, according to Computerworld.
That matters because it changes the job AI is being asked to do. Prompt-based systems can draft a memo, summarize a meeting, or generate code snippets. Agentic systems are supposed to go further. They are designed to take a goal, break it into steps, and work through the sequence with less back-and-forth from the user. In OpenAI’s telling, ChatGPT Work is built to automate knowledge work across business applications while maintaining enterprise-grade governance and security, Computerworld reported.
The practical appeal is obvious. For enterprise teams, the promise is not just faster drafting. It is software that can move from a request to a finished output across files and tools, which is a different category of product altogether.
What ChatGPT Work is designed to do
OpenAI’s description of ChatGPT Work points to a system meant for workplace execution, not just conversation. Computerworld said the platform can coordinate multiple tools, operate across applications and files, and generate business documents, presentations, spreadsheets, and websites.
That range is important. It suggests the product is being aimed at the routine, cross-functional tasks that fill modern office life: preparing a presentation from a set of notes, turning a spreadsheet into a report, assembling a first-pass website, or organizing information spread across different files. Those are the kinds of assignments that often require a human to move between applications, copy material from one place to another, and keep track of what comes next.
If ChatGPT Work can reduce that friction, it could become useful not only for individual employees but for automation teams looking to offload repetitive knowledge work. The broader pitch is that AI no longer sits at the edge of a workflow as a helper. It becomes a participant inside the workflow itself.
That is also where the enterprise questions begin. OpenAI says the platform is designed with governance and security in mind, but the source material does not spell out which controls customers get or how those controls work in practice. For IT and security teams, that missing detail will matter as much as the feature list.
GPT-5.6 is being sold on economics as much as capability
ChatGPT Work arrives alongside a broader rollout of GPT-5.6, which OpenAI says offers stronger performance across coding, enterprise knowledge work, cybersecurity, and scientific research while also reducing inference costs and token consumption, Computerworld reported. The models are now generally available through ChatGPT, Codex, and the OpenAI API, according to the same report.
The most notable part of that message may be the economics. OpenAI said it trained GPT-5.6 “to get more useful work from every token,” and described the result as “stronger performance per dollar,” meaning either more output for the same spend or similar output at a lower total cost, Computerworld reported.
That framing reflects where enterprise buyers are now. Many organizations are no longer evaluating AI models only on capability or benchmark scores. They are watching token use, inference bills, and whether a workload is worth running on a premium model at all. Counterpoint Research vice president Neil Shah told Computerworld that rising token consumption has created “bill shocks” for enterprises and is pushing companies to use different models for different workloads.
In other words, model quality still matters. But price per task matters too. A system that performs well and consumes fewer tokens has a better shot at winning enterprise budgets.
The rollout also signals a more segmented enterprise strategy
OpenAI is not just releasing one product and one model. It is creating a layered set of offerings. Computerworld reported that OpenAI priced Sol at $5 per million input tokens and $30 per million output tokens, while Terra and Luna are lower-cost options for organizations scaling AI deployments.
The source material does not explain the technical differences among those tiers, but the strategy is clear enough. OpenAI appears to be segmenting its enterprise lineup so customers can choose between cost and capability based on the task at hand. That makes sense in a market where not every workflow needs the most expensive model.
It also suggests OpenAI is thinking beyond pure model access. By offering ChatGPT, Codex, the OpenAI API, and now ChatGPT Work with multiple pricing paths, the company is building a more complete enterprise stack. For buyers, that could simplify procurement. For competitors, it raises the bar.
Why agentic software matters for day-to-day office work
The case for agentic enterprise software is straightforward: a lot of office work is still made up of small, repeated steps. People pull data from one system, rewrite it in another, create documents, build slides, update spreadsheets, and package the result for someone else. None of those tasks is especially glamorous. All of them take time.
That is why the shift from prompt-based assistance to delegated task execution matters. If AI can reliably handle the handoffs between applications and files, it can start to look less like a writing aid and more like an operational tool. That is a bigger promise, but also a riskier one. The more work a system carries out on its own, the more important accuracy, auditability, and security become.
The government backdrop to GPT-5.6’s rollout adds to that tension. Computerworld said the models reached general availability weeks after a limited preview that followed U.S. government restrictions tied to advanced cybersecurity and biology capabilities. The report does not say which restrictions applied, or whether any remain in place, but the history is a reminder that enterprise AI is not being deployed in a vacuum.
The bigger enterprise AI question
ChatGPT Work is best understood as a sign of where enterprise AI is headed. The category is moving away from one-shot generation and toward systems that can complete work across tools, files, and applications. OpenAI is betting that businesses will see value in that shift, especially if it can deliver better performance per dollar.
That bet is plausible because it matches how enterprise buyers already think. They want automation that saves time, but they also want predictable cost, manageable risk, and enough control to satisfy security teams. ChatGPT Work does not answer every one of those concerns yet. The available reporting leaves important gaps around governance, integrations, and the practical differences among OpenAI’s pricing tiers.
Still, the direction is hard to miss. OpenAI is not just asking enterprises to chat with software. It is asking them to delegate work to it.