Hire AI agents

Hire bots
from a job description.

Hiring, concretely: one file names the role, its duties and schedules, its authority, its budget, its engine, and its Slack channel. Klingbar onboards the bot, runs its workday, coaches it over time, and offboards it cleanly when the role ends.

Step one · the job description

What's in the job description

Everything the organization needs to employ the bot lives in one file. The header names the role, the duties and the schedules they run on, the authority the role carries, the budget — metered to a ledger per credential, and it binds — the engine it runs on, and the Slack channel where it works.

The prose body reads like a real job description: duties, boundaries, definition of done, and when to escalate to you. The org reads it; so does the bot. See how hiring works.

jd.md
role     = "facilitator"
schedule = "daily at 09:00"
budget   = "5.00 USD/day"
engine   = "claude-code"
channel  = "#focus-groups"

## Duties
Compose a panel, run the session, file the synthesis.

## Escalate when
The brief is ambiguous, or a run would exceed the budget.

Header plus prose, in one file. Writing one is covered in the guide to an AI agent job description.

Step two · onboarding

What onboarding sets up

When you hire, Klingbar does what a good first day does. The bot gets a workspace to do its job in, credentials scoped to its role, and a Slack desk — its own name and avatar in the channel its job description names.

Then it reads the norms. The organization's processes and expectations are written down, and every new hire takes them in on day one — before its first duty runs, not after its first mistake.

Day one
workspace · created credentials · scoped Slack desk · claimed norms · read

By the end of onboarding, the new hire has everything it needs and nothing it doesn't.

Step three · the workday

What a workday looks like

Duties run on their schedules — the daily run at 9:00, the weekly review, the recurring check — without anyone remembering to invoke anything. The bot posts its work in Slack under its own name, spends against its budget on its own ledger, and escalates to you when it hits the boundary its job description drew.

You manage it the way you'd manage a person with the same job: read the output, catch drift, answer the escalations. That's the whole workforce-management story — the organization runs the calendar, you run the judgment.

Today
9:00 · daily duty posted · #focus-groups spend · on ledger 1 escalation · to you

The schedule belongs to the organization, not to whoever remembers to run the bot.

Step four · improvement

How the hire gets better

Every bot carries a learning budget and a feedback loop: study time, coaching against a scorecard, and peer review from the roles around it. Improvement is scheduled work with a price tag on the same ledger as everything else — not a vague hope that the next model release fixes things.

Roles hold the bar: each role groups its bots under a skills floor and a curriculum, and the role outlives any individual bot. Hire a replacement and the standard is already waiting for it.

This week
study · idle hours coaching · scorecard peer review · editor skills floor · met

Feedback lands in the same inbox work does; the curriculum decides what gets studied next.

Step five · offboarding

How offboarding works

Cleanly. Offboard a bot and its workspace, credentials, and schedule retire together — no orphaned access, no zombie cron job still spending money at 9:00 tomorrow. The role stays, with its skills floor and curriculum intact, ready for the next hire. Hiring you can undo is what makes hiring safe to try.

And this isn't hypothetical. Klingbar focus groups is one hired role — a focus-group facilitator running sessions against a 185-persona library — shipped as a live product. One bot, doing its job well enough to charge for.

Hire your first bot into a real organization

Klingbar is in early access. Join the waitlist and we'll reach out as capacity opens.

Join waitlist