ChatGPT Has Started Responding When Something Happens in Your Business
One of the things I keep noticing with AI is that the really useful changes are not always the ones with the biggest announcement around them.
Sometimes a fairly small product update changes how you could actually work with the technology.
OpenAI has added webhook-triggered tasks to ChatGPT Business. In fairly ordinary language, that means ChatGPT can now respond because something has happened somewhere else.
A new Gmail message arrives.
Somebody posts feedback in a Slack channel.
A GitHub pull request changes.
Instead of somebody noticing that event, opening ChatGPT and deciding what to ask it to do, a task can already be sitting there waiting for it. OpenAI gives examples including summarising a new email, drafting next steps after client feedback appears in Slack and reviewing changes to a pull request.
I think this matters because most of the way we have used ChatGPT until now has still been surprisingly manual.
Even when people talk about AI saving time, the process often starts with a human remembering to use it.
An email arrives and you copy it into ChatGPT.
A client sends feedback and you ask ChatGPT to analyse it.
You remember on Friday morning that you want a weekly report, so you open a conversation and start gathering things together.
Scheduled tasks began changing that because ChatGPT could do something at a particular time without you having to initiate it every time.
Webhook-triggered tasks take that another step.
The trigger can now be the work itself.
I can see why this sounds rather technical, particularly if you have never had any reason to care what a webhook is. You do not actually need to understand the plumbing to understand why the change matters.
A webhook is essentially a way for one system to tell another system that something has happened.
The more interesting question is what happens when ChatGPT can then respond intelligently to that event.
For a small expert-led business, this begins to change where AI sits in the workflow.
It stops being quite so dependent on somebody thinking, I should ask ChatGPT about this.
You can start deciding in advance what should happen when particular situations occur.
That is a different way of designing the business.
Imagine you regularly receive enquiries about speaking.
You might want the incoming message summarised, the sender and event identified, the important details extracted and a first assessment made against your own speaking criteria.
Or perhaps you have client feedback arriving in Slack.
Rather than somebody collecting that feedback at the end of the month, a defined task could respond as it appears and begin organising what needs attention.
There are plenty of fairly mundane examples like this, and I actually think the mundane ones are where the value often sits.
You do not necessarily need AI doing something spectacular.
You need it removing the little interruptions and repeated decisions that keep pulling you away from the work you actually need to do.
There is another part of today's release that I think deserves attention.
ChatGPT Business users can now share scheduled and webhook-triggered tasks with other people in their workspace.
Each person gets an independent copy. They can change the instructions, connect their own apps and use their own permissions.
That starts to make an AI workflow more like a business asset.
If I create a particularly good process for reviewing an incoming opportunity, I do not necessarily have to explain the process to somebody else and hope they recreate it accurately. The workflow itself can be shared.
That is something I have been thinking about for quite a while through the work I do with AI agents.
The valuable part is often not the prompt.
It is the thinking that has gone into deciding what should happen.
What information needs to be considered?
What standards should be applied?
What should the AI do when it is uncertain?
What needs a human decision?
When should it stop?
Those are governance questions, and they become more important once the AI can begin acting because an event has occurred.
There is an obvious connection here to authors as well.
A book business produces events all the time.
Reader emails arrive. Reviews appear. Speaking enquiries come in. Podcast invitations arrive. Beta-reader comments are collected. Launch data changes. Somebody asks for support.
A lot of that information currently sits in separate places and waits for the author to notice it.
The possibility of an AI system beginning to work around those events is interesting.
For AI Book Companion™, I think the immediate opportunity is around the ecosystem surrounding the Companion rather than trying to automate the reader's actual conversation with the book.
If feedback about a Companion arrives, it could be gathered and classified.
If readers repeatedly ask for help with the same point, that information could eventually become useful to the author.
If a support question comes in that sits outside the Companion's permitted scope, a workflow could make sure it reaches the right place.
The boundary matters here.
I would be very wary of automating everything simply because the technology makes it possible.
There are parts of an author-reader relationship where somebody needs to know that a human has actually looked at what they said. There are decisions where context matters too much to hand over. There are also obvious questions around privacy, permissions and what information an AI workflow should be allowed to access.
OpenAI's implementation at least starts from individual connected-app permissions, which means shared tasks do not automatically give everybody access to somebody else's Gmail or Slack data. Each user connects their own account and operates through their own permissions.
I think that distinction will matter more as these systems become normal.
The interesting part of AI automation is becoming less about how many things we can automate.
It is increasingly about deciding which events deserve an automatic response, what that response should be allowed to do, and where human judgement still belongs.
Today's OpenAI update is quite a practical little example of that shift.
ChatGPT is beginning to wait for the work to happen.
And once AI can do that reliably, the way we design our businesses around it changes.
Three Key Insights
1. ChatGPT can now be triggered by events rather than only by people or schedules.
ChatGPT Business supports webhook-triggered tasks for new Gmail messages, Slack channel activity and GitHub pull-request activity.
2. AI workflows are becoming shareable business assets.
Scheduled and webhook-triggered tasks can now be shared across a ChatGPT Business workspace, with each user creating their own copy and using their own connected-app permissions.
3. Governance becomes more important when AI starts responding automatically.
The useful question is no longer only what ChatGPT can do. Businesses increasingly need to define which events should trigger AI, what it should be allowed to do afterwards and which decisions remain human.
Three Questions & Answers
What is a webhook-triggered ChatGPT task?
It is a task that begins when a supported event happens in another application rather than at a fixed time or after somebody manually starts it. OpenAI currently lists new Gmail messages, Slack messages and GitHub pull-request activity as supported triggers for ChatGPT Business.
Does sharing a task give somebody access to my Gmail or other connected apps?
No. OpenAI says recipients create their own copy of the task and connect their own apps. The task then operates using that person's permissions.
Why does this matter to a one-person expert business?
Because many small pieces of work begin with an event: a client replies, an enquiry arrives, feedback is submitted or something changes in a project. If the first useful AI action can happen automatically when that event occurs, there is less need to repeatedly notice, transfer and initiate the work manually.
Sources
OpenAI — ChatGPT Business Release Notes, 25 August 2026
https://help.openai.com/en/articles/11391654-chatgpt-business-release-notes
OpenAI — ChatGPT Enterprise & Edu Release Notes, 25 August 2026
https://help.openai.com/en/articles/10128477-chatgpt-enterprise-edu-release-notes
I use AI to help research, organise and refine my writing. The ideas, expertise and intellectual property are my own, and every article is reviewed and approved by me before publication.