If Google can’t make AI assistants useful, maybe no one can

For years, tech companies have promised that AI will give everyone a personal assistant but have only offered something more akin to an experienced human. In the last six months, this has started to change, thanks in large part to OpenClaw’s OpenClaw software. And among the top AI labs chasing similar success, one seems poised to create assistants on a large scale: Google.

At I/O 2026, Google announced new AI assistants to gather information, schedule events, summarize your inbox and calendar, and more. Agents can run in the background continuously, and the company says it will integrate with Google and external devices. It’s also expanding its design tools and revamping Search with additional AI features. Some are rolling out this week, and some will be available in the coming months, but the company’s strategy seems clear: take some of the things that have helped OpenClaw succeed and expand it with Google’s deep knowledge of our digital presence.

“Before that, I think AI agents were more experienced in research,” said Koray Kavukcuoglu, CTO of Google DeepMind and Google’s chief AI architect. Seaside in conversation. This year, he hopes, he will be “in our life.”

AI agents have been buzzing since ChatGPT was launched in late 2022, but it remained a science fiction concept until the rise of OpenClaw, which has gained millions of users since its launch last November. OpenClaw allows people to communicate with their agents through daily apps such as WhatsApp and Telegram, and (as long as the laptop is open) agents can run during the day. They were good enough to perform the necessary tasks reliably, though some logical errors.

It made all the AI ​​labs sit up and take notice, but OpenAI was one of the first players to take action, acquiring OpenClaw (although it remains open source) in February and hiring its creator Peter Steinberger. But Google’s existing services offer a huge advantage. Where OpenClaw managed parenting by integrating with tools people already use, Google can do the same through MCPand create deep links to its internal products, including Gmail, Drive, Docs, Photos, and Search. If anything, it’s surprising that it took this long.

One of Google’s biggest bets this year is Gemini Spark, its new AI assistant for consumers. Google has promised that Gemini Spark can work on all Google services and more than 30 external partners are coming soon, including Dropbox, Uber, and Spotify. Gemini Spark and clouds; it can run 24/7 without opening a laptop and can sync across the web, Android, and iOS. The assistant is going to trusted testers this week, and the beta will be available in the US next week on Google Ultra plans.

Google likes to use Gemini Spark, like shopping, searching, and coordinating with other people’s schedules and plans. Google also hopes that people will find something to use. Josh Woodward, head of Google’s Gemini program, says he’s been using Gemini Spark to organize a neighborhood party, send delegates to follow up on RSVPs and what attendees bring, send reminders, and see when the homeowner’s association allows to put in a giant. Outside of Spark, Google is also introducing Daily Brief, a morning update similar to OpenAI ChatGPT Pulse.

Gemini Spark isn’t available yet, but if it works the way Google says it will, it could be a big step forward for the AI ​​industry. Google’s oldest experiment completed the work at a snail’s pace while hacking your browser. Author last year’s Gemini 3 releasehis assistants were successful at some tasks – such as cleaning out inboxes – but still failed at others. Now, Google is taking a step that can help with mirroring others important things of OpenClaw: long-term assistants who work day and night in the background, giving them the ability to have information about their work – and giving users the ability to text and email their assistants directly.

Starting this summer, Google’s AI search, too, is getting helpers – and promises that it will eventually do more than eat up public buildings and make recommendations. pizza and glue. “Information agents” need to do constant background research – like tracking market trends or the weather for the perfect day for a picnic.

Google also announced the expansion of Antigravity, the development platform it launched six months ago. The new Antigravity desktop application will serve as a central support for agents, and the entire system is now designed as a platform for building and managing independent agents, Google says. The development follows similar tools from OpenAI and Anthropic, which have worked to expand their successful coding services into accessible tools for non-programmers.

All this will be supported by a new series: Gemini 3.5, whose first entry Gemini 3.5 Flash should be available next month. The prototype is supposed to have a much better recording capability than the Gemini 3, which it was release from great joy Last November. It aims to integrate changes from Anthropic, which is known for its technical expertise, and OpenAI. Gemini 3.5 Flash is especially good for “deploying multiple agents at the same time and completing long-term tasks,” Kavukcuoglu told reporters on Monday. It should also be four times faster than other types of borders and less than half (or in some cases, one third of) the cost – a fact that is suitable for 24/7 AI agents, and the cost of presentation that only adds up quickly.

In the world of AI agents, Google still plays with the one-man team behind OpenClaw. But it is a long-time leader in AI competition, and its software has a big advantage: It now supports more than 900 million users per month, executives told reporters on Monday, in more than 230 countries and more than 70 languages. Compared to dedicated AI companies under economic pressureit can offer temporary subsidies to attract users. And although its representatives have not met the real world, they are going in the right direction. If any AI company can make assistants useful, it’s Google. If it can’t, it doesn’t have much reason to go back – and the whole idea may need to be reconsidered.