Gemini’s new AI assistant is a Google feature

Google’s new “24/7” AI assistant, Gemini Spark, can be surprisingly good at doing things for you. But I’m not sure it’s worth the money and potential privacy trade-offs.

The company gave me an opportunity to go to Spark last week. Google advertises Spark as an AI assistant that can handle tasks and tasks in the background — even tasks that involve multiple steps — that allow you to put down your phone or step away from your computer. It also advertises very high Spark website that “it’s always under your control,” that “you choose to turn it on,” and that it’s “designed to check in with you before you do big things.” Given the growing skepticism of AI, it’s very important that “My ‘not participating in artificial AI’ T-shirt has people asking questions that have already been answered by my shirt.”

I didn’t know where to start, so I took a page my friend Antonio’s book: I decided to use Spark to do something like what Google showed on stage at I/O. Would it work as well in my home office as it did on the big stage?

At I/O, Google Vice President Josh Woodward showed off a few different models. The first was to ask Spark to write an email to the Google team, write everything about the launch of Gemini Live and “last week’s success,” and use special AI skills to make the email sound like him. Google is asking Google to do Google things they should being the easiest carrier in the world, so I tried to push it further.

I asked Gemini to write an email to my wife that collected all of our monthly purchases in 2026. I thought this test would tell me a few things: Can Spark know who my wife was (without giving her Spark’s name), can it know where our budget spreadsheet is in Drive (which doesn’t have “budget” in the file name), and can it write an email in Gmail?

When I got the results from Spark shortly after, I said: “Wow, that’s nuts.” Spark found my wife’s email address, extracted the correct information from our 2026 budget page, took monthly purchases including outstanding items from May (which had not yet expired when I tested), calculated them all, and put them all into an email in my Gmail. The text of the email listed my wife’s first name, even though her email does not have her name. It also included a leave sign that we only use for each other.

In his next example, Woodward asked for help planning a block party. I am not planning a block party, but I asked Spark for help using the questions they asked. It didn’t go well. It created a table for friends and family as “a real clue of who’s bringing what,” wrote an email in my Gmail mentioning a shared page that never existed, and created a terrible train with detailed photos of city permits.

To push Spark, I asked it to create the missing signature sheet and add a link to the email that was already written. While Spark took a few minutes to figure it out, the process worked; created a spreadsheet and went back to the email address and put in the link.

Woodward’s final performance was undoubtedly the most impressive. He spoke to Spark to ask him to do a lot of things: make his meetings with CEO Sundar Photosi hot pink on his calendar, write a note to a new neighbor to invite him to his block party, and create a document to support his children’s activities at the end of the school year. For my brand, I asked them to make a calendar event every month before my wife’s birthday and make it pink, and email my family to send them the first episode of the latest season. TaskmasterI am creating a document with the top things that my wife and I need to know to prepare our son for school.

I started this request at 3:35PM PT on Friday. During I/O, Woodward hinted a bit about putting his phone down and promised to check the results later in the main event, which he did. But when I spoke with one problem – Spark wanted to connect with those I declined, which I declined – my work was done four minutes later.

Again, I was a little impressed with the results, even if they weren’t perfect:

Spark can be a powerful tool. But there are a few caveats I should mention. Like all AI tools, you still have to check the output to make sure it’s correct, which can have high parameters when it comes from personal data to prepare things that you share with people you know. Although Google markets Spark as something that can work by itself, I found myself constantly looking at it or looking at the notifications it sends to my phone. What good is an agent if you have to monitor everything they do instead of trusting them? And why should I have any problem with not being sure about draining energy from a resource-hungry storage space for unnecessary tasks?

Currently, Spark is available to subscribers of Google’s AI Ultra plan, which starts at $99.99 per month, and only to US users and in English. Google gave me the opportunity to test Spark, and I don’t think it’s good enough to be the only reason for those expensive plans. Especially when I can do all the tasks I asked Spark to do on my own – it just takes more time.

Spark also works well if you are already deep in the Google ecosystem and have Personal Intelligence. I’ve had a Google account for almost two decades, so Spark has a lot to draw on to inform its solutions. But when Google they promise that Gemini doesn’t “learn directly” from your Gmail inbox and with Personal Intelligence turned on, you have to trust Google that it will good steward of your data. At this point, I’m not sure if this is cheap or dangerous.