The Household Is the Unit

In #014 I made a claim I want to hold you to: the person who buys trust when the user can't audit the agent is not the user. It's the household-side person. The adult child who set up the assistant. The household manager who configured it. The parent who carries the regret.

I ended that edition with a forward-pointer. I promised you the commercial pattern: who actually writes the check, what the buy decision looks like when buyer and user are different people, and what that means for pricing and distribution at scale.

This is that edition. And it turns out the news handed me the perfect anchor.

The News: Shopify Noticed the Agent. It Hasn't Met the Buyer Yet.

On June 17, 2026, Shopify published its Spring '26 Edition: 150+ updates built entirely around agentic commerce infrastructure. The numbers are real and staggering. AI-driven traffic to Shopify stores grew 8x year over year in Q1 2026. Orders from AI-powered searches were up nearly 13x. New buyers placing first orders through AI channels are converting at nearly twice the rate of other channels.

Shopify's read on this is correct as far as it goes. They are building the checkout layer for wherever AI conversations happen. Catalog feeds the recommendation. Checkout completes the transaction. Agentic Storefronts put merchant inventory inside every AI surface that matters.

Here is what the entire architecture assumes, in a sentence buried in their own explainer:

"A shopper might ask an AI chat, 'Find me a birthday gift for a 10-year-old who likes art.' The AI agent searches across connected product catalogs, considers what it already knows about the shopper's preferences and budget, and presents a curated set of recommendations."

One person. One conversation. One budget. Shopper = agent operator = buyer = user.

That's the assumption. And for a meaningful segment of the market, the segment that powers Shopify's 13x order growth, it's correct.

But there's a second segment that is just as large, and structurally different. And every pricing, distribution, and trust model being built for agentic commerce has the wrong unit of analysis for it.

The unit isn't the user. It's the household.

The Assumption I Carried Into This Arc (And Where It's Evolved)

When I started the ICP migration arc in #010, I was updating my view of who the ultimate Accords buyer actually is. I was moving from enterprise security teams toward consumer households. I was looking at the boomer parent, the family IT person who became the default tech support for people they love.

Here is where my model was incomplete.

I was still thinking of the household as a collection of individual users. I was asking: which individual in the household is our buyer? The answer I kept landing on was the household manager, the person who handles the family's digital life.

That framing is wrong in a specific way.

The household isn't a collection of individual users who happen to share an address. It's a commercial unit with a shared budget, joint approval authority, and a non-uniform distribution of technical fluency. The decision to deploy an AI agent into a household is not an individual decision. It's a household-level governance call, even when only one person makes it.

The buy decision is made at the household level. The usage is distributed across users who did not make that decision. The trust is delegated by the buyer to the agent, which then represents the buyer's authority in front of people who never consented to that contract directly.

This is a completely different commercial structure than a subscription the user buys for themselves. And nobody is pricing it that way.

How the Buy Decision Actually Works

Let me be specific about the mechanics.

A household-level AI purchase has a recognizable structure. It shows up across enough scenarios that the pattern is real.

One person holds the financial relationship with the vendor. They are the account holder, the person on the receipt, the person who gets the renewal email. Call them the account principal.

That person configures the agent. They set the preferences, the spending limits, the scope of what it can touch. They do this once, imperfectly, and then mostly don't revisit it.

One or more other people in the household use the configured agent. They interact with it daily. They may not know what it is, what it can do, or what limits exist. They trust it because someone they trust told them to.

The account principal is not present at the moment of most transactions. They are not watching their elderly parent ask the agent to rebook a flight. They are not monitoring their spouse's interaction with the agent managing household subscriptions. They set the ceiling and then they left the room.

The commercial question this creates is: who is the vendor's customer?

Right now, the answer in every pricing model I've seen is: the account principal. The one paying. The one on the subscription.

That is the right answer for pricing. It is the wrong answer for product design, distribution, and trust architecture.

What the Household Unit Changes About Pricing

If the household is the unit, then the per-seat pricing model, the thing every SaaS company defaulted to for twenty years, has a serious problem.

Per-seat pricing assumes each seat represents a roughly equivalent user with roughly equivalent needs. That's true in enterprise software. It is not true in a household.

In a household, one person makes the purchase decision and bears the accountability. Multiple people benefit at asymmetric utility levels. The account principal gets governance value, they're buying peace of mind, control, and the ability to extend their care across people they're responsible for. The dependent users get the surface-level benefit: the thing just works and someone set it up for them.

These are not the same product. They have different value propositions, different willingness to pay, and different success metrics.

Every AI subscription priced at a flat $20/month per user is leaving the table set wrong. The household manager who deploys an agent across three family members and a dependent parent is not buying four $20/month subscriptions. They are buying one governance deployment with four usage surfaces. The value they are capturing, at the principal level, is not linear with seat count. It scales with the number of people they can extend care to without being physically present.

The pricing model that reflects this is not per-seat. It's per-household, with principal-level governance features as the upsell, not usage limits.

Nobody is selling this yet. The pricing infrastructure for agentic commerce is being built around transaction volume and catalog access, Shopify's model, not around the governance layer the principal needs. That gap is a product category waiting to be filled.

What It Changes About Distribution

The distribution implication is harder to see but more important.

If the buyer is the account principal and not the user, then every acquisition channel aimed at the end user is going upstream in the wrong direction.

Consumer AI adoption looks viral because users share things they like. They tell their friends. They post screenshots. The thing spreads user to user.

But for the household-level purchase, the buyer didn't decide because a friend who uses it told them it was great. The buyer decided because they needed to solve a problem that was their responsibility to solve. An elderly parent who needed help. A teenager who needed guardrails. A spouse who needed their digital life simplified without having to care about how.

The buyer's decision is driven by obligation and care, not curiosity and delight.

That means the acquisition channel for the household buyer is not the user-side content loop. It's the channel that reaches people in the moment they realize they have a responsibility they can't discharge with current tools. That moment is specific. It's a care decision, not a discovery moment.

Distribution that works for individual users, product virality, influencer seeding, developer communities, will reach the household buyer by accident at low rates. Distribution that works for household buyers runs through caregiving networks, financial planning contexts, and the kind of content that reaches people who are trying to figure out how to be responsible for someone else's digital life.

Most consumer AI companies are distributing to the wrong person.

Where Accords Sits In This

I've said it once already this arc and I'll say it one more time.

Accords governs what agents do on the open internet.

The household commercial unit creates a specific problem for that governance layer. The account principal deploys once. The agent then acts across every surface the user touches, across Shopify's checkout, across appointment booking, across financial accounts, across everything Apple's routing layer hands it on the other side of the edge.

The governance has to travel with every one of those actions. It cannot live in the account principal's settings console. It cannot be a log they check afterward. It has to be present at runtime, at every transaction, for every user the principal is responsible for, whether the principal is watching or not.

That is not a feature of the AI agent. It is a separate infrastructure layer. The agent capability and the principal's mandate travel together or the commercial model breaks. The principal's willingness to pay, the value that makes the household-level subscription meaningful, evaporates the first time the agent acts outside the ceiling they set, and they find out from a receipt instead of an alert.

The consumer surface where this reaches the account principal is AtHand. The governance promise isn't abstract. Help is always AtHand, and the people you're responsible for are covered the way you would cover them yourself.

The Falsifiable Close

Here is what would break the household-unit thesis.

If the major AI platforms converge on a family plan that correctly prices governance-at-the-principal-level, distributes through caregiving and financial-responsibility channels, and includes a portable mandate that travels with the agent across all surfaces, without needing independent infrastructure, then the commercial gap I'm describing is a transitional window, not a structural one.

Watch for: any major platform launching a "household" or "family" subscription tier that explicitly separates principal-level governance from dependent-user access. If that product ships with spending controls, plain-English receipts, and scope constraints that hold outside the platform's own ecosystem, the infrastructure argument weakens significantly.

As of June 23, 2026, no platform has launched that product. The $20/month per-user model remains the template. Shopify's Spring '26 Edition, for all 150 updates, has no concept of a household principal separate from a transacting user.

The gap is still open.

Next week: the permission architecture question this raises. When the account principal delegates authority to an agent that acts on behalf of a dependent user, what is the legal and commercial structure of that delegation? Who is liable when the ceiling fails? What does the contract actually say? #016 will be about the governance document that doesn't exist yet.

If you're forwarding this, send it to whoever in your life is managing someone else's digital infrastructure without a framework for it.

Max