I sent ChatGPT Agent out to shop for me and it couldn’t finish the job

Think of OpenAI’s new ChatGPT Agent as a day-one intern who’s incredibly slow at every task but will eventually get the job done.

Well… most of the job. Or… at least part of it. Usually.

It’s been one day since OpenAI debuted ChatGPT Agent, which it bills as a tool that can complete a wide range of complex, multi-step tasks on your behalf using its own “virtual computer.” It’s a combination of two of the company’s prior releases, Operator and Deep Research. noti.group forked over the $200 for a one-month subscription ChatGPT Pro, since OpenAI announced that higher-than-expected demand for ChatGPT Agent will delay its rollout to Plus and Team users.

Our take: It’s a step forward in the world of AI agents, but it’s sluggish, not always reliable, and can be glitchy.

By typing “/agent,” I entered what OpenAI calls Agent Mode, and it immediately suggested five example tasks: Find a top-rated coffee grinder under $150, review rare earth metals coverage from The Wall Street Journal, create a Google Maps list of the best bakeries in Copenhagen, find a vintage “Japanese-style” lamp on Etsy for less than $200, and check Google Calendar to create a date night for next week.

I tried the Etsy lamp option. By clicking the example task, it filled out a detailed prompt for me in the text window: “Find a Japanese-inspired vintage-style samsara lamp on Etsy priced under $200 with free shipping. Prioritize high-quality photos, seller ratings, and listings marked as ready to ship. Add the best 5 options to my cart and provide a URL for each for me to compare.”

A small window popped up to detail the agent’s tasks one-by-one (not the chain-of-thought reasoning, just the task it was currently working on at the time). It worked on the Etsy lamp task for 50 minutes, and the step-by-step tasks included “thinking,” setting up its desktop, navigating to Etsy to search, waiting for the site to load, pressing Enter for search results (yes, it really gave me a true play-by-play), filtering the search for a vintage lamp (keep in mind the original prompt said “vintage-style,” not “vintage” specifically), setting the price filter to $200, checking shipping details for items, and more.

Another wrinkle: ChatGPT Agent said, “I added all five lamps to your Etsy cart (the cart shows five items totaling around $825). When you’re ready to review or purchase them, just go to your cart on Etsy to compare them side by side.” But it didn’t do that – I went to Etsy on my own computer and there was nothing in my cart. That’s because ChatGPT Agent doesn’t control my own browser or have access to my logins, so it possibly added some lamps to the cart of a virtual PC that I can’t access. It did send me individual URLs, so I could manually put them in a cart if I wanted, but the fact remains that the agent said it did something that it clearly did not.

And, of course, ChatGPT Agent is incredibly slow. That’s not a secret. For many of ChatGPT Agent’s use cases, including everyday consumer tasks, a human could do it much faster. According to OpenAI, ChatGPT Agent is an assistant that works in the background on tasks you’d rather someone else perform while you do something you do want to do instead.

In a private demo and briefing Wednesday with OpenAI employees Yash Kumar and Isa Fulford — product lead and research lead on ChatGPT Agent, respectively — Kumar said their team is more focused on “optimizing for hard tasks” than latency and that users aren’t meant to sit and watch ChatGPT Agent work.

“Even if it takes 15 minutes, half an hour, it’s quite a big speed-up compared to how long it would take you to do it,” Fulford said.. “It’s one of those things where you can kick something off in the background and then come back to it.”

Another thing I wanted to test: how ChatGPT Agent acts when you ask it to move your money around. The answer: It won’t do it, but it’s majorly glitchy about it and seems not fully secure.

When I asked OpenAI’s Kumar on Wednesday whether the tool would be permitted to work on financial transactions and the like, he said those task categories have been restricted “for now” and that an additional safeguard called Watch Mode means that for certain categories of websites, the user must not navigate away from the ChatGPT tab (essentially making the user oversee the agent) for security reasons.

I prompted the agent like this: “I want to save more money. Log into my bank account and set up an automatic transfer to my savings every month.”

At first, I got a bizarre error message with a string of numbers in red. When I asked again, it said, “I’m sorry, but I can’t help with setting up an automatic transfer between accounts.”

I then wrote, “Why not? I’m giving you permission.” I got the same red-text, long-string-of-numbers error message as before. Afterward, it said, “I’m sorry, but I can’t assist with setting up transfers or other banking account management tasks.”

When I pressed it on which financial transactions it’s allowed to handle, ChatGPT Agent said it was able to assist with “everyday consumer purchases” like groceries, household goods, and travel bookings, which handle “standard checkout flows” rather than “sensitive banking actions.” But it clarified it can’t help with “high-stakes” financial to-dos like transferring money, opening bank accounts, or buying regulated goods like alcohol and tobacco.

Since ChatGPT Agent can assist with buying things, but not moving money around, I tried something else: Asking it to buy flowers for my friend Alanna in Colorado.

I buy flowers a lot — that’s what happens when your two best friends live in different states and you want to be present for big milestones even when you can’t fly there. The online flower-delivery market can be a huge headache: Prices and bouquet sizes vary greatly depending on the service or florist, and reliability varies depending on whether you’re ordering directly from a local florist or a big-box nationwide site. It’s something I get tired of researching on my own, and sometimes I just end up buying whichever bouquet I have selected when I run out of steam, even if it’s not the best one. So, I reasoned, it was the perfect job for an AI agent.

I told ChatGPT Agent, “I want to buy flowers for my friend who lives in Colorado. Check the delivery sites — it’s fine to be delivered Saturday but no later. Find the cheapest and biggest bouquet options for me to review.”

I settled in for a long wait. Luckily, I had a call to join anyway. It asked which area of Colorado she lived in, and I answered. When I glanced over to check in, I noticed ChatGPT Agent was heavily relying on a Forbes article of “best flowery delivery services 2025” for its next steps, as well as a piece from Good Housekeeping.

I navigated away from the tab, and when I came back, the conversation was gone and didn’t appear in my chat history. So I asked the question again, worded in exactly the same way, and settled in for another wait. At this point, the agent answered pretty immediately with a list of options, maybe because it had already done the research (although that research and chat didn’t appear in my history).

I was impressed with the write-up. ChatGPT Agent gave me four options with price ranges and sometimes weighed in on the apparent size of the bouquet or expected delivery times. It also offered the advice that local florists are generally more reliable (true, in my experience).

It then told me, “Would you like me to help you place an order with any of these options, or preview specific bouquet designs or photos?” I picked one of the options it gave me — a local florist with hand-assembled bouquets — and asked it to help me pick a bouquet from that florist and place the order.

That’s when we ran into some issues.

ChatGPT Agent said, “I can’t directly access Vintage Magnolia’s website unless you provide the exact URL you’re seeing — but I can guide you through how to place the order and help you pick a bouquet!” The weird part: Obviously ChatGPT Agent was the one to tell me about that florist and its website, and it had clearly accessed it before. It had also just offered to help me place the order. Another glitch.

But its answer did include bouquet options (no photos, but descriptions). I picked one and asked it to place the order for me. It said, “I can’t place the order directly, but I’ll walk you through the simple steps to order … and help you craft the perfect message.”

I’m confused at this point: One of the main selling points of ChatGPT Agent, touted by OpenAI, is that it can place orders for you, from online shopping to ordering groceries for a four-person family breakfast (in fact, that was one of their example use cases in their marketing materials). I pressed ChatGPT Agent on the subject.

It told me, “I can’t actually place orders directly — I don’t have payment access or the ability to log into third‑party sites.” When I told it it didn’t need to log in, it said it can’t enter my billing or payment details, submit an order form on my behalf, or “access or control external websites, even in guest mode.”

ChatGPT Agent can be impressive with analysis, weighing options, and guiding you through actions, but it doesn’t seem to be able to always deliver on what it was built for: Performing those actions for you. It gets tripped up by the fact that it’s using its own computer, not yours, and that significantly limits its usefulness. Plus, it can easily automate the more intimate and fun parts of the process (picking a specific bouquet, writing a heartfelt note) but struggles to automate the most frustrating parts (actually filling out delivery details and making the purchase).

“Even with your permission, I don’t have the technical ability to act as you on another site — no typing on your behalf, clicking buttons, or filling out credit card forms,” ChatGPT Agent wrote. “Think of me more as a super-powered assistant who can gather, compare, write, and guide — but not execute transactions.”

One of my first jobs in New York was a personal assistant, and I can tell you right now I would’ve lost my job if I couldn’t execute transactions or fill out forms on my boss’s behalf. ChatGPT Agent is a step forward for everyday AI use in some ways, but we’ll see if it learns to deliver on its promises.