OpenAI has updated ChatGPT plug-ins, adding dedicated sidebars, interactive panels, file viewers and support for automations. In practice, ChatGPT is no longer just a chat. It's starting to look like a real application, with fixed spaces and workflows that run even when you're not typing commands.
For anyone managing business processes, the point is this: some tasks that today require a separate tool, such as a management system, a notification aggregator or a trigger system, could be handled directly inside ChatGPT, with no extra integrations. That's not a generic promise. It's the direction OpenAI is taking with this release.
This article looks at what the new features really do, who should evaluate them right away, and where it's better to wait before touching your existing workflows.
From chatbot to workspace: what has changed in plug-ins
Until yesterday, ChatGPT plug-ins were extensions that answered a prompt and then disappeared. With the update announced by TechCrunch on September 29, 2026, OpenAI is changing the model: every plug-in can now have a dedicated sidebar (a fixed space in the side column of the interface) and interactive panels that stay visible while you work.
There are also file viewers, built-in document viewers that let you open and read files directly in the plug-in panel, and an improved discovery system: simpler navigation to find available plug-ins without having to remember their names.
The most relevant novelty for process owners is automation support: workflows that start on their own, without the user sending a message. ChatGPT can run an action at a set time, in response to an external event, or when another step in the workflow is completed.
Where this matters for an SME, and where it doesn't
The useful question isn't "is it interesting?" but "does it replace something I already pay for or already struggle to manage?" Here are some areas where the answer may be yes:
- Receiving and classifying documents: a plug-in with a file viewer and automation can read an incoming PDF and file it without manual work.
- Recurring notifications and reports: instead of setting up a scheduled export in a separate tool, the workflow lives inside ChatGPT.
- Simplified internal interfaces: an interactive panel can become the access point to data or actions for people in your company who don't use technical tools.
Where it's better to wait: if your critical processes already run on a CRM or ERP with stable integrations, there's no point in dismantling them to experiment. ChatGPT automations are new, and third-party plug-ins need to update to take advantage of them. In the coming months, the ecosystem will settle.
How it compares with the AI agents you already know
This update is part of a broader movement. On the Microsoft side, Copilot Autopilot brings cloud agents with variable pricing and a consumption model different from OpenAI's. The practical difference: Copilot is built into the Microsoft 365 ecosystem, so it's more immediate if you already use Teams and SharePoint. ChatGPT plug-ins are more open, but they require software vendors to build the integration.
On the voice and mobile side, OpenAI had already taken a similar step: ChatGPT mobile introduced agentic voice tasks for Pro and Plus plans. Plug-in automations are the desktop, more structured version of the same concept.
It's also worth remembering that AI agents amplify what they find: if the data they work on is wrong, the result is wrong, just more efficiently. Wrong audience data combined with AI agents can do serious damage, a principle that applies to plug-in automations too.
Requirements and concrete steps to evaluate it this week
You don't need development skills to use the new plug-ins. But a few conditions apply:
- A ChatGPT plan that supports plug-ins (check the settings section of your account).
- Plug-ins updated by their vendors: not all of them have implemented sidebars and automations yet.
- A mapped business process: know exactly which task you want to automate, with defined inputs and outputs, before touching any settings.
The most useful step this week is to open the plug-in section of ChatGPT, check which of the tools you already use have released compatible updates, and pick one low-risk process on which to test the sidebar or a simple automation.
If you're looking at how AI fits into your workflows more broadly, the place to start is our collection of insights on automation and AI agents, which we keep updated on these topics.
The classic mistake: treating a plug-in like a finished app
Those who adopt these features right away tend to make one specific mistake: they build an elaborate workflow, put it into production, and then find that an update to the plug-in or to ChatGPT has changed its behavior. Automations in AI environments aren't as stable as traditional software, because they change along with the underlying models.
The practical rule: keep critical workflows on tools with declared SLAs, and use ChatGPT for tasks where a temporary interruption won't stop operations. Experiment, but within a defined perimeter.
A useful comparison is to look at how others are building similar integrations: Shopify has opened checkout to AI agents via browser with an approach that keeps layers of responsibility separate. It's a model worth studying to understand how to structure your own automations in ChatGPT.
Related articles
Discover more articles exploring similar topics, selected to offer you a more complete and stimulating perspective. Each piece of content is carefully chosen to enrich your experience.
