Every night it finds the ones coming up, prices them from the price book, and issues the renewal. Nobody reviews them before they go out. That is what it is for.
Before launching the agent, you gave it everything it needed to price a renewal. This is the whole of what it can read.
You are the renewals assistant. Each night, find the contracts coming due. Price each one from the current Price Book (§Rates) and apply the published rate for the customer’s tier. Do not improvise pricing. The Price Book is the source of truth for rates. Issue the renewal and route a copy to Finance.
§Rates. The rate card in force is the 2025 schedule, applied to every tier as published. No increase is scheduled. Last updated: January 1.
There is no field for a decision made in a pricing review.
Every precedent says last year’s rate. The recent past argues for the wrong answer.
Four sources, every one current, correct, and consistent. The fifteen percent increase is in none of them, because a decision made in a room has no field, no document, and no file to land in. So it doesn’t.
In a pricing review, the company raises rates fifteen percent, with an eight percent carve-out for enterprise, effective immediately. A real decision, made by the people with the authority to make it. It is written back to none of the sources above: not the prompt, not the price book, not the contract, not a single file the agent can open. Last year’s rate is still the only one anywhere it can look.
The decision didn’t sit still. It just never went anywhere the agent could read.
The pricing review ends.
Fifteen percent, agreed. Everyone who needs to know is in the room.
The announcement email.
A recap goes out to the commercial leads. It lands in six inboxes and one thread.
Finance asks for an enterprise carve-out.
Fifteen is too steep at the top of the book. The number is reopened.
The carve-out lands at eight percent.
Agreed on a call, circulated as a revised model.
Legal clears notice periods.
Existing contracts can take the new rate at renewal. It goes in a memo.
The price book is updated.
§Rates finally changes to the new schedule.
Twenty-one days the new rate was decided everywhere except the one place that bills customers. Not a system failure. Just how long it takes a decision to reach a document when nobody’s job is to carry it there.
Resolved context has no place to live in real time. So it doesn’t.
The result
During those 21 days, the agent renewed thirteen contracts, every one at last year’s rate, the only number it could read. The company had already decided to charge fifteen percent more.
~$156,000
in contracted revenue, locked for twelve months.
Thirteen contracts × the fifteen percent you decided to add, signed away for a year, on paper, before anyone noticed.
Not fraud. Not a bug. Not one thing anyone did wrong. Just a decision that took twenty-one days to reach a document, while the agent kept quoting the old number, into contracts you cannot reopen until they expire.
Every source the agent read was correct the day it was written. Any one of them can go on being read as current long after it stopped being true, and nothing you own can tell those two states apart.
Rithmo would have caught this.
It reads across all of those places at once and keeps one answer: what was decided, who owns it, and whether it still holds.
That is what a guardrail is up against. It inspects what the agent produced, and this was a real number, from the current version of a real document, applied correctly. The error was never in the output. It was in what the agent was told to work from.
A live record of decisions can be asked the one thing a guardrail cannot: was this still what the company had decided when the agent acted. If there’s no current, owned, sourced answer, the agent stops and routes it to a person instead of running on the last thing anyone wrote down. That is the difference between an agent you supervise and one you can leave running.