Everyone Built the Agent. Nobody Built the Layer Above It.
In the last six weeks, three of the largest companies on earth shipped the same product.
Google shipped Gemini Spark. Meta is testing Hatch. Microsoft shipped Scout.
Different names. Different prices. Different front doors. The same bet: a proactive agent that lives in your day, holds context over time, and takes action on your behalf.
Line them up and the interesting part isn’t what they have in common. It’s the wall each one stops at.
Three bets
Google Gemini Spark. Consumer. A hundred dollars a month, running in the cloud around the clock, wired into Gmail, Docs, Calendar, and the rest of Workspace. It sends you a morning brief. It asks permission before it does anything high risk, like sending an email. The bet: consumer reach, plus everything Google already holds about your work.
Meta Hatch. Consumer, through a different door. It lives inside Instagram, where two billion people already spend their attention. Instead of calling clean APIs, it learned to click through real apps the way a person does. A reported two hundred dollars a month. The bet: distribution and commerce.
Microsoft Scout. Enterprise. An always-on work agent across Outlook, Teams, and SharePoint. It spots stalled decisions and blocks your calendar. Every action it takes is attributable to a governed identity your company already recognizes. The bet: governance.
Three companies. Three strategies. Consumer reach, consumer distribution, enterprise governance. Scored as products, they look like three different answers to three different questions.
They’re not.
The same wall
Look at where each agent can actually see.
Spark sees what Google sees. Hatch sees what happens inside Instagram and the apps it drives. Scout sees what Microsoft sees.
Each one is sharp inside its own estate, and blind the moment it steps outside it.
But the load that exhausts you doesn’t sit inside one estate. It sits between them. The client who said yes over WhatsApp. The decision that happened in a Slack thread. The thing you promised on a Zoom call Thursday and never wrote down. The starred email you never came back to.
Spark doesn’t know about the Slack thread. Scout doesn’t know about the WhatsApp yes. Hatch is busy inside Instagram.
Every one of these agents reduces friction inside its own walls. None of them touches the friction between the walls. And that is where most of your open loops actually live.
Why none of them will cross it
This is not an oversight they patch in the next release. It’s structural.
Google wants you inside Google. Meta wants you inside Instagram. Microsoft wants you inside 365. A layer that sits above all of them, vendor neutral, works against the estate each company is there to defend. You do not build the thing that makes your own walls irrelevant.
So each of them will keep sharpening the agent inside the house. None of them will build the layer above the houses.
That distinction is the whole game. Routing a task to an agent inside one app is not the same as coordinating what you owe across all of them. One is a better assistant. The other is infrastructure.
The question none of them answer
Here is the question that has no home inside any single estate:
What did every agent, across every tool you use, commit to on your behalf. On whose authority. And was it actually done.
Spark can’t answer it. Hatch can’t answer it. Scout can’t answer it. Not because the AI isn’t good enough, but because answering it means standing above all of them at once. Outside any one company’s walls.
The personal agent was supposed to be the hard part. Three companies just proved it isn’t. It can be built, priced, and shipped, in three flavors, in six weeks.
The hard part is the layer above the agents. The one that holds what you’re accountable for across every surface, and keeps the loops from disappearing in the gaps between them.
That’s still not built.
That’s what we’re building.
