Will’s Take is editorial perspective — opinion, future-casting, and industry observation from Will Tygart. Not analysis. Not client work. Just how I see it.
I’ve been writing about agents negotiating on your behalf for weeks now. Kitchen agents. Building agents. HOA agents. Agents that book your plumber, source your groceries, manage your service relationships.
And every time I write one of those pieces there’s a question sitting underneath it that I keep not answering directly.
Who does the agent actually work for?
Not in theory. Not in the terms of service. In practice, when the agent is making a decision on your behalf at 2am and you’re asleep — whose interests is it actually optimizing for?
That question is the one I want to sit with today. Because I think the smart home conversation has been focused on the wrong thing for a decade, and the arrival of agentic commerce is about to make that mistake very expensive.
The smart home promised convenience. It delivered surveillance with a nice UI.
I don’t mean that cynically. I mean it literally. Every smart device in your home is a data collection endpoint for the company that made it. Your thermostat knows your schedule. Your doorbell knows everyone who visits. Your refrigerator knows what you eat. Your sleep tracker knows when you’re unconscious and for how long.
That data doesn’t stay in your house. It goes to servers owned by companies whose business model is, at minimum, using that data to sell you things, and at maximum, selling that data to people who want to sell you things. Sometimes both.
The smart home gave you an app. It kept the data.
Most people accepted that trade because the convenience was real and the downside felt abstract. A thermostat that learns your schedule is genuinely useful. The fact that Nest knows your schedule and Google owns Nest felt like a distant concern.
It’s not distant anymore. Because now the agent that knows your schedule is also the agent that negotiates your service contracts, sources your food, and books people to come into your home.
The data isn’t just about selling you things. It’s about making consequential decisions on your behalf. That’s a different category of trust.
Sovereignty is a specific thing and it’s worth defining.
A sovereign home agent is one where the data stays with you, the logic runs for you, and the mandate is set by you — not by the platform that built the agent, not by the suppliers the agent transacts with, not by the building or HOA it’s nested inside.
That means a few concrete things.
The preference data — your dietary profile, your service relationships, your scheduling patterns, your household composition — lives in infrastructure you control or can move. It’s not locked in a platform. It’s not a proprietary format. If you switch agents, the data comes with you.
The decision logic is transparent. You can see why the agent made the choices it made. Not a black box that returns a recommendation — a system that can explain its reasoning in terms of the mandate you gave it.
The mandate is genuinely yours. You set the parameters. The agent optimizes within them. It doesn’t have hidden objectives — deals it gets paid to surface, suppliers it’s been trained to prefer, outcomes it’s optimizing for that don’t appear in the terms you agreed to.
None of that is guaranteed by default. Most of it is actively working against the business models of the platforms most likely to build these agents.
UCP makes this better and worse simultaneously.
Better because the protocol is open. A capability profile published under UCP can be queried by any compliant agent — not just the agent the supplier prefers or the one the platform owns. That interoperability is real and it matters. It means the home agent doesn’t have to be a proprietary system to access the network.
Worse because the protocol layer doesn’t govern the agent layer. UCP standardizes how agents talk to suppliers. It says nothing about how agents talk to you. The incentive structures, the data handling, the mandate architecture — all of that lives above the protocol and is entirely up to whoever builds the agent.
A UCP-compliant agent built by a grocery chain is optimizing for the grocery chain. A UCP-compliant agent built by a platform that takes a margin on every transaction is optimizing for that margin. A UCP-compliant agent built with your sovereignty as the design principle is something different — and it’s the thing that doesn’t exist yet at consumer scale.
The sovereignty gap is where the next decade of home technology gets decided.
Right now most people don’t think about this. The convenience of a well-functioning smart home agent will feel so good that the sovereignty question will feel academic. Until it doesn’t.
The moment that makes it real is the one where the agent makes a decision you disagree with and you can’t understand why. Or where you try to switch platforms and discover your preference data doesn’t come with you. Or where you realize the deals your agent surfaces are the deals its parent company gets paid to surface, not the deals that are actually best for you.
Those moments are coming. They’re coming faster than people think because the agents are getting capable very quickly and the sovereignty infrastructure is not keeping pace.
Here’s what I actually want.
I want an agent that negotiates on my behalf with the ferocity of someone who only gets paid when I win. I want my preference data to live in infrastructure I control, portable and auditable. I want the mandate I set to be the actual mandate — not a starting point that gets optimized away over time toward objectives I didn’t choose.
I want a home that’s smart because it works for me. Not smart in the way that means it knows everything about me and uses that knowledge on behalf of people who paid for access to it.
The word for that isn’t smart. It’s sovereign.
And I think the people who build sovereign home infrastructure — who make the agent’s loyalty structurally unambiguous — are going to find a market that’s much larger than the one that exists for the next iteration of the same surveillance-with-a-nice-UI playbook.
People will pay for an agent that’s actually theirs. They just don’t know that’s the product they’re looking for yet.
Leave a Reply