In May 2026, Google announced roughly 100 AI features over two days at Google I/O. Many of the journalists covering the event couldn't keep up with all of it, and among tech analysts the phrase "I/O spaghetti" made the rounds — features sprawling in so many directions that it was hard to tell what the center was.

This isn't unique to Google. The same week, Microsoft, OpenAI, and Meta each unveiled new AI features of their own, and as the competition accelerates, people trying to adopt AI tools have settled on a common refrain: "I don't know what to use." The gap has widened between how fast new features arrive and how fast they can be folded into real work. The more features there are, the heavier the burden of choosing becomes. Google I/O was the biggest stage yet on which that played out.

What Google Did This Time

At I/O 2026, Google chose to embed AI across its existing services rather than launch new products.

AI Overviews now sit at the top of the search page by default. Type a question and you get a summarized answer before any list of links. Gmail gained email draft suggestions and automatic sorting of incoming mail. Google Photos analyzes the background and context of a photo and attaches a caption automatically. Chrome bundles and summarizes the contents of your open tabs, and route guidance in Google Maps has turned into a conversational question-and-answer experience. YouTube connected AI to video summaries and comment replies.

There were features for developers, too. Jules, a coding agent, connects to GitHub repositories to find bugs automatically and propose fixes. Veo 3, a video generation model, produces high-definition video clips from nothing more than a short text description. NotebookLM expanded its ability to analyze multiple documents at once and quickly locate specific content.

Google DeepMind also unveiled its research direction on "World Models" — technology that lets AI understand the causal relationships and spatial structure of the physical world, as if running a simulation. It's a long-term bet tied to fields like robotics, medical diagnostics, and climate forecasting, but at the announcement stage it was unclear which products it would land in, or when.

There's a logic to this strategy. Build a separate app and users have to learn it from scratch. Put AI inside the tools people already use every day and adoption climbs with no learning curve. It's the same approach Microsoft took by tucking Copilot inside Word and Excel. The more the AI gets used, the more usage data accumulates, and that data feeds the training of the next model. For a platform company, the longer this loop runs, the bigger the competitive advantage.

But the Strategy Draws Criticism, Too

There is plainly a skeptical view of this announcement.

Embed AI into everything and the polish of each feature becomes uneven. Users can't tell what actually works well until they try it themselves. With roughly 100 features, finding the ones worth using becomes its own time cost. And when a feature behaves differently than expected, the disappointment rarely stops at that one feature — it can shake trust in the surrounding services as a whole.

Tension inside DeepMind surfaces in reports from time to time. DeepMind was originally an independent organization focused on AI safety research and long-term alignment problems. Since being folded into Google, it has been expected to contribute to commercial product development as well, and from the outside it isn't clear where the real priority lies between safety research and revenue. Some researchers warn that commercial pressure could crowd out long-term safety work. How fast to move — between rushing World Models technology into products and verifying its safety carefully — will be a point of contention going forward.

One more thing: putting AI everywhere is a platform's competitive strategy. Trying every AI becomes a practitioner's way of spending time. Just because the strategy is rational for the platform doesn't mean the same strategy is right for the practitioner. Lose sight of that difference, and your work cycle keeps getting set to the platform's release calendar.

What This Leaves for Korea's One-Person Businesses

Some of what was announced at Google I/O can go to work right now.

Gmail's draft suggestions can cut time spent on repetitive email. If you regularly exchange similar messages with the same clients or partners, editing an AI-generated draft is faster than writing from scratch. NotebookLM's expanded features are for getting through long contracts, proposals, and market research reports quickly — finding a specific clause or a key figure without reading the whole document. Note organization tied into Google Docs reduces the time it takes to structure what came out of a meeting.

By contrast, World Models research, Google Beam — the holographic video-conferencing system — and complex AI agent frameworks are not at a stage a solo business owner can actually adopt in 2026. Roughly half of the announcements were previews of where things are headed one to two years from now.

The problem is the pressure that builds right after the event to try everything immediately. Social feeds fill up with feature roundups, and dozens of I/O recap videos land on YouTube. React to that noise by pouring energy into exploring new tools, and the work that was running smoothly slows down. The pattern of AI launch news disrupting work routines now repeats every quarter.

Something similar happened when smartphones took off. When hundreds of thousands of apps poured out, people felt they had to try them all. But look at who actually became more productive, and they were using fewer apps, not more. The person who used two apps deeply worked faster than the person who used five. The practitioners who survived wave after wave of fast-changing smart technology had one thing in common — and it wasn't being first to adopt. They knew where the bottleneck was in their own workflow, and they picked only the tools that fit that spot.

The Practitioner's Call After I/O

There are things you can try at work right away, now that I/O is over.

Write down the AI tools you're currently using. Split them into the ones you've actually opened in the past two weeks and the ones you haven't. The tools you haven't opened probably aren't needed right now. The work got done without them.

Pick the one repetitive task that eats the most of your week. Email triage, research, drafting reports — anything. Spend just three days finding a way to handle that one task with AI. That's less time than it takes to read every article about Google I/O. If it works, the tool earns a place in your workflow. If it doesn't, you've learned that this one feature, out of the hundred or so announced, doesn't fit the way you work.

How Google folds DeepMind's World Models research into its products, and how DeepMind balances safety research against commercial goals, will become clearer in announcements to come. For now, those questions matter less than figuring out where AI fits in your own work.

On the day Google ships a hundred features, a practitioner will pick two or three of them — and six months from now, the person who got better at the one tool they already use will likely be getting more done than the person who followed every I/O announcement.