Can't my users just use ChatGPT?
Why an AI agent inside your product and ChatGPT in its own tab do different jobs, and how to tell which one your users need.
Quick answer
ChatGPT is excellent at helping the person typing into it, in its own window, with what it can see. An agent inside your product does something different: it works where your users already are, signed in as them, and carries out the steps in your product through your own API, asking before it changes anything. Use ChatGPT, or Claude, to build and set things up yourself; give your users an agent in the product when they need work done there.
A bias to declare: I build Verb, an AI agent that goes inside software products. I use ChatGPT and Claude every day, including to build Verb, so this is not a post about either being bad.
A commenter on Reddit put it plainly: the setup steps for an in-app agent looked like something they already get by prompting free Claude or ChatGPT. It is a fair question, and the answer is not about which model is smarter. It is about where the work happens and whose account it happens in.
What ChatGPT is great at
ChatGPT helps the person typing into it. Explaining a concept, drafting an email, writing code, planning a migration: it is superb at all of it, in its own window, with whatever you paste in or it can find on the web. For you, the person building a product, it is one of the best tools there is.
What it cannot see from there
Now picture one of your users, signed in to your product, trying to set up a recurring report for their team. If they open ChatGPT and ask how, it can only answer from what is public: your help pages, if it finds them. It does not know their account, their plan, their data or which screen they are on, and it cannot press anything in your product for them. The best it can give them is a list of steps to follow back in the other tab.
That is the moment users get stuck in the first place. A list of steps is what they already had in your docs.
What an agent inside your product does instead
An in-app agent lives where the user already is. It knows who is signed in, because your app tells it. When the user asks for the recurring report, it does the steps for them through your product's own API, as that user, so it can only do what they could do by hand. Before it changes anything, it shows them exactly what it is about to do and waits for a yes.
What about a ChatGPT app?
You can build an app or connector so ChatGPT can reach your product. It is the right move if your users spend their working day inside ChatGPT. Most do not: they work in your product, which is where they get stuck, so that is where help gets used. Asking them to leave, install your app in ChatGPT and sign in again adds the friction you were trying to remove.
Where ChatGPT and Claude do come in
Setting the agent up. The steps the Reddit commenter saw, describing your product's actions and wiring in the script, are exactly the kind of work a coding agent does well. Verb has an MCP server for this, so Claude Code, Codex, Cursor or ChatGPT can set the agent up from your codebase: see Verb for AI agents. Your coding agent does the setup; your users get the agent.
The verdict
- You, building or running the product: ChatGPT or Claude.
- Your users, stuck on something they want done in your product: an agent inside it, acting as them.
- Your users who live in ChatGPT all day: a ChatGPT app may be worth it.
If you are deciding what kind of AI your product needs at all, how to add AI to an app you already have starts a step earlier.
Frequently asked questions
Can ChatGPT do things inside my app for my users?
Not by itself. It answers in its own window from what it can read. To act in your product it needs a connection you build, such as an app or connector, and the user has to leave your product, install it in ChatGPT and sign in there.
Isn't an in-app AI agent just ChatGPT with my docs?
No. A model answering from your docs explains how to do something. An in-app agent does it: it calls your product's API as the signed-in user, so the change happens in their account, after they say yes.
Should I build my product a ChatGPT app instead?
Only if your users spend their day in ChatGPT. Most use your product to do their work, and that is where they get stuck, so that is where help is used.
Can I use ChatGPT or Claude to set up an in-app agent?
Yes, and it is the quickest way. Verb's MCP server lets Claude Code, Codex, Cursor or ChatGPT set the agent up for you; it is integrated in under an hour in most cases.
Vishnu, founder of AskVerb @itsvishnups
Building the AI agent that lives inside a SaaS product and does the work its users ask for. Writing down what I learn along the way, including the parts I got wrong first.