Disarmed by default

Disarmed is the starting position

Every role on EmployLabs starts disarmed. Nothing is sourced, sent or spent until you arm the role with a preset and caps. A role you create today sits there, ready, and contacts nobody, because nothing runs unattended until you deliberately say so.

That is an unusual default for a product whose entire argument is that the work can run without you. It is deliberate. The value of an autonomous recruiter is not that it acts; it is that it acts the way you would have. Those are different claims, and you should be able to settle which one is true before the first message leaves.

Nothing runs until you arm it, and arming it means naming the limits.

Arming is a decision with limits attached

You arm a role by choosing a preset and setting caps. The preset decides how much of the pipeline is allowed to run. The caps decide how far it may go. Neither is implied by creating the role, and neither can be skipped, so there is no moment at which a new role quietly begins to work on your behalf.

Arming is also per role. A search you know well can be armed while the one you are still shaping stays disarmed, and turning a role’s automation off again is a switch on that role rather than a setting buried somewhere else.

Confirm mode: asking before each send

Armed does not have to mean unattended. Per-stage settings let outreach and scheduling run in confirm mode, which asks before each send. The engine still does the finding, the screening and the drafting; you see what it proposes to send and decide whether it goes.

• Automatic. The stage runs to your caps without asking.
• Confirm. The stage prepares the work and waits for your yes before each send.
• Off. The stage does not run at all, and the work stays with you.

Why the burden sits with us

There is a way of introducing automation that asks for faith first and evidence afterwards: switch it on, watch what happens, judge it on the results. It is a poor deal for the person being asked, because the cost of being wrong lands on their domain, their candidates and their reputation, and none of it can be taken back.

We think a product confident in its own judgement should start from a position where it cannot act until it is told how far it may go. Disarmed-by-default is that position made into the default rather than a setting you have to find.

Learn it on roles you know

The useful way to evaluate this is not an abstract one. Take a role you have worked yourself, where you already know who the plausible candidates are and what a bad approach looks like. Arm it with outreach in confirm mode and read each proposed message as you would a junior colleague’s draft: would I have contacted this person, and is this how I would have put it.

Disagreements at that stage are cheap and informative. They are usually a brief that says less than you thought it did, or a screening bar set somewhere you would not have set it, and both are adjustable. Disagreements found after four hundred emails have gone out are neither cheap nor adjustable.

Confirm is a posture, not a permanent queue

Confirm mode is for the period in which you are forming a view, and forming a view is supposed to end. Once the proposals stop surprising you, move that stage to automatic and let it run to its caps. A review queue you must clear item by item for ever is not automation, and nobody should have to live in one.

How to use it

1. Start on a role you know well rather than the one you are least sure about. You are testing the engine’s judgement, and that requires you to have judgement of your own to compare it with.
2. Arm the role with a preset and conservative caps, and set outreach and scheduling to confirm.
3. Fix what you disagree with in the brief and keep going until the proposals contain nothing that surprises you. That is the moment to move a stage to automatic, and not before.