Set a budget, and never spend past it

Autonomy and spend are the same conversation

Anything that contacts candidates and interviews them on its own is spending money on its own. That is the whole promise, and it is also the reason most teams hesitate. A system that works while you sleep can also spend while you sleep, and the failure mode people imagine is waking up to a bill for a search that was never going to work.

So the cap is not a settings-page detail. It is the thing that makes the autonomy usable.

A ceiling per role

You can set a ceiling on what a role may consume. When it is reached, the paid work on that role stops (sourcing, outreach and interviews). Nothing spends past it, and nothing continues quietly at a reduced rate.

Being told no is free. Nothing is bought in order to discover the limit.

That second part matters more than it sounds. The naive way to enforce a cap is to do the work, then check the balance, which means every refusal still costs you the thing it refused. Every check on the platform happens above the paid call, so a role that has hit its ceiling stops without spending anything to find out.

What each action costs, before you run it

Sourcing itself costs you nothing — you can search the market as widely as you like. What draws from a balance is the work that produces something: reaching a candidate, getting a candidate's answer, revealing a contact on its own, and interviewing them. Each is visible, with a ledger you can reconcile line by line. The point is not the accounting. It is that you can answer “what did this role cost us” without an invoice arriving three weeks later to tell you.

Nothing autonomous starts on its own

• Every role begins fully manual. Automation runs only on roles you have explicitly armed, at the level you chose.
• Arming requires a decision, not a default.A role that has not chosen how far to automate cannot be armed at all, because “on” without a posture means something different to every person who reads it.
• Approving a send or a spend refuses an agent caller. Meera can recommend the entire configuration and cannot arm it. That refusal lives in the code rather than in a prompt.

When something stops, it says why

A stopped pipeline with no explanation is worse than one that keeps running, because you cannot tell it apart from a broken one. Every stop carries a sentence naming what stopped, why, and what would restart it. If the cap is what stopped it, look at what it bought and what the pool looks like now, so raising it is a decision rather than a reflex.

How to use it

1. Set the role ceiling at the point you arm automation. It is a limit rather than a budget, so set it to the most you would be willing to spend, not the amount you expect.
2. Read the stop reason before you raise the cap. Sometimes the honest answer is that the pool is telling you the brief is wrong, and more sourcing deepens the problem.
3. Reconcile against the ledger at the end of a role. Cost per role is the number that tells you which searches are worth automating.