I tried Google's free AI agent for handling everyday chores

Wait 5 sec.

Google has widened access to the agentic capabilities of Gemini through Tasks, the renamed Spark feature. The AI can carry out jobs across websites and connected apps for any user, an appealing proposition for anyone whose to-do list contains a few things that really should have been done last Tuesday.The important difference between Tasks and regular Gemini is what happens after the chatbot gives you an answer. Ordinary Gemini can recommend a restaurant or help compare products, and its existing connections to Google services already make it useful for finding information. Using Tasks means it can actually go to different websites and complete multiple steps of a process beyond simply putting together information.Gemini Tasks is now available to free users in six countries, including the US, Canada and Australia, but UK users will have to wait for the free rollout or check whether their paid Gemini subscription provides access.If you're in a country where Tasks is available, you can try it by switching from Chat to Task in a Gemini conversation, then giving it permission for different activities and services as it works. You can intervene when necessary, and Gemini will request approval before taking any actions that might require extra approval on your part, like submitting your information to an external website. Check out some of the ways it can be useful that I tested below. 1. Curing the car(Image credit: Shutterstock)For my first test, I asked Gemini for help getting my car an oil change and tire rotation. I asked Gemini through Tasks to find a service center near me that could handle both jobs during the same appointment and to find a time in my schedule for it. The AI found two locations and suggested one would be more convenient as it had reviews mentioning getting everything done in one visit. It suggested a couple of times that looked free on my schedule, and it took care of making the reservation for me when I gave my approval. It wasn't necessary here, but Tasks can even call businesses to make appointments in some cases. 2. Shop along(Image credit: Shutterstock)Next, I used Tasks for some shopping. On a whim, I asked for something small: a few basic charging cables I needed. But I asked it to do technical comparisons and find the best option that could power my laptop and phone simultaneously and had a discount from a good brand. I was especially interested in whether it could distinguish between a genuinely good deal and a cheaper charger that didn't meet my requirements. A single-port model might cost less, but that wouldn't help much when I wanted to charge two devices at once. The AI came back with two good options. Notably, while regular Gemini can compare specifications and recommend a charger, Tasks can continue into retailer websites, investigate current offers, and work through checkout after receiving the necessary approval. It can look for a better price and prepare a purchase rather than leaving me to repeat the search on a retailer's website.3. Returning the goodsMy next ask was almost the inverse. I picked a recent purchase I made online, a sweater, and asked the AI to help me return it. Gemini searched through my Gmail, found the sweater and its details, and worked out if it was still eligible for a refund. It visited the website and worked out the return policy, which in this case was to refund the money to my card.Regular Gemini could help me locate the email and explain the retailer's return policy. But using Tasks meant it could connect the process of looking through my email with opening the retailer's website and finding the relevant order, then working through whatever return procedure the company has devised. Gemini prepared the form for me, then asked if I wanted to submit it. I held off for now, but it was good to know I could speed up the process if necessary this holiday season. Of course, until Google gives me a robot capable of walking to the post office, that part remains my responsibility. 4. Dinner plans(Image credit: Shutterstock / Mila Naumova)Choosing a restaurant can become a surprisingly involved process, but Google promises that mentions of using Tasks can speed things up. I asked Gemini to find and make reservations for a nearby restaurant with good menioning comfortable seating and vegetarian options on Saturday around 7 p.m.I was particularly interested in whether it could keep looking when the first restaurant didn't have a suitable time. There's little point in an agent finding the perfect restaurant if the only available table is at 4:15 in the afternoon. I wanted something reasonably close to my preferred time, with enough flexibility to consider another option if necessary.Regular Gemini can already make restaurant recommendations and help compare menus. But Tasks meant the AI, with my permission, would make the reservation, navigating the booking system as easily as the review pages. But in about 10 minutes, I not only had a reservation, but had a few extra reservations for next month set up as well.5. Tracking down wandering packages(Image credit: Future)For my final experiment, I turned to the Tasks feature to find all shipping confirmations from the previous two weeks, see when my packages were expected to arrive, and flag anything that appeared delayed.While I could have already used it to search connected Gmail messages for tracking numbers and summarize the information in the emails found, Tasks took the extra step of opening carrier websites and checking the current status of each shipment. That matters when the original shipping confirmation might be several days old, and the information in it may no longer reflect what's happening. I asked it to put everything into a short summary and suggest what to do about anything that seemed overdue. The two packages used different delivery services, but the AI visited both pages: one showed the package out for delivery, while the other had been stuck at "Shipping Label Created" for several days. Gemini suggested contacting the company I bought from and drafted an email for me to approve, which I did. It's not glamorous, but it was remarkably efficient. The first message to send the email took about four minutes.This might be the least glamorous use for an AI agent imaginable, which is probably why I find it appealing. I already know how to track a package, and I don't need Gemini to explain the process. I'd much rather have it check the tracking pages itself and tell me whether there's anything worth worrying about.None of these services are particularly complex, but they are necessary and often annoying enough to encourage procrastination. Almost anyone with a browser and enough patience could complete them. The attraction of Tasks is that it can potentially do the clicking and navigating that ordinary Gemini generally leaves to the user.Of course, an agent working through a retailer's checkout or a restaurant's reservation system has more opportunities to make mistakes than a chatbot simply recommending somewhere to eat. But you can keep an eye on anything involving money or a confirmed booking.I might have finished the jobs faster myself, too. But the convenience of just sending the agent off to do the work for me is nice, especially when it involves more than one multi-step process. I don't particularly need Gemini to tell me how to return a sweater, but I'm delighted it can handle most of the process while I do something else. And since it's now available to everyone, a lot more people are likely to get comfortable with the idea of AI agents and how useful they can be, even with the most mundane jobs. If Gemini can keep handling those small jobs reliably without constant supervision, I might finally run out of excuses for leaving them unfinished.