High usage is not the same as safe revenue.
Hollow Usage is when usage rises while the value behind a renewal quietly falls. The dashboards look healthy. Underneath, no one is checking whether the usage still means anything.
Usage is easy to measure. So it becomes the stand-in for health. Real renewal value is harder. It lives in outcomes: a champion who still cares, a result the buyer can name. When teams scale, they watch the easy number and lose the hard one.
A renewal held up by a usage chart is fragile. Usage can stay flat while the reason to pay erodes. When the champion leaves or the budget tightens, the number does not hold. Gross retention is where this shows up, usually a quarter too late to fix.
Usage stops standing in for value. The product ties usage to the outcome it should produce. It flags when the two come apart. A human still owns the account. The product tells them when a healthy account is thinning out.
Everyone already agrees on the setup. Agents have no habit and no loyalty, so switching costs fall and the old lock-in goes with them.
The argument the industry cannot settle is what replaces the moat. Some say data. Some say trust. Some say the system of record. Some say usage itself. Every one of those has merit. Every one of them misses the same thing.
The popular take is that switching costs are dying, so usage becomes the new moat. The first half is true. The second half is not.
Usage holds value only when a human is attached to it. Take the human away and you have volume without loyalty. Agent traffic can climb every month and still mean nothing at renewal, because no one in the account would miss you if you were gone.
Volume is not a moat. Attachment is. When the two decouple, usage stops predicting retention, and the SaaS metrics you report to the board start lying to you. This is why the "usage is the new moat" thesis is only half right. It counts the activity and ignores whether anyone would defend it.
The moat was never only the data or the screens. It was a person.
Somewhere in the account, a human used your product every day. They built their work around it. They could not picture doing the job without it. They trained new hires on it. They defended the line item when finance came hunting. They noticed when the team outgrew the plan and asked to buy more.
You did not pay them. They did your value capture for free.
An agent does none of that. An agent runs the task and moves on. It does not build a habit. It does not grow attached. It will never walk into a budget meeting and fight for your renewal.
So when the agent becomes the user, something worse than churn happens. The champion does not just leave the account. The champion never forms in the first place. There is no one to lose, because there was never anyone to win.
This is the mechanism behind Borrowed NRR and Expansion Orphans: revenue that looks retained or expandable on paper but has no human owner underneath it. Hollow Usage is the upstream cause. Kill the champion and the whole renewal-and-expansion engine loses its driver.
The fix is not to keep AI agents out. They are coming, and your customers want them. The fix is to design for value capture after the agent does the work, so the value stays visible, defensible, and commercial.
Agent usage only creates durable value if a human can still see it, defend it, and pay for it. Real usage holds all three. Lose even one and the usage goes hollow.
Run every AI-touched workflow through three questions.
1. Visible. When the agent finishes, can a human still see what the product did? You need real proof, shown when someone needs it: manual reviews avoided, a renewal risk caught weeks early, a workflow closed days faster. A champion cannot defend value they cannot see.
2. Defensible. Does the product own something an agent cannot get cheaper elsewhere? It might own the trusted record, the audit trail, the permission model, the proprietary data, or judgment that would be costly to rebuild. If the same outcome is one tool away, the agent will take it.
3. Commercial. Does the value map to how the customer buys, renews, and expands? The real question is whether you charge for what the agent does or for the outcome the customer would defend in a budget meeting. Price the outcome, and usage and revenue grow together.
Usually, yes. The risk is treating usage as proof of value when it is only proof of activity. The two move together, until they don't.
Hollow Usage is one renewal resting on activity instead of outcome. Borrowed NRR is the portfolio version, where the retention number hides weakening economics.
Hollow Usage is when product usage keeps rising while the value that drives renewal falls. It is common in accounts where AI agents drive the activity and no human champion forms behind it, so the dashboard looks healthy while retention, expansion, and pricing power quietly erode.
Usage only holds value when a human is attached to it. Agents have no habit and no loyalty, so agent traffic can climb every month and still mean nothing at renewal. Volume without attachment is a moat on a chart and a commodity in the room.
It weakens revenue without shrinking it. Renewals lose their internal defender, expansion signals go dark because no one notices the team is ready to buy more, and pricing power softens. An acquirer sees the difference even when the topline does not: usage goes up, the multiple comes down.
Run each AI-touched workflow through the Hollow Usage Test. Ask whether a human can still see what the product did (visible), whether the product owns something an agent cannot get cheaper elsewhere (defensible), and whether the value maps to how the customer buys and renews (commercial). Lose any one and the usage is hollow.
Seven questions. Five minutes. A pattern read on the spot, no call to see it.