An unfillable role stays open for months before anyone questions the brief
The role is created by bringing together everything the team wants. That produces a seniority level the budget cannot afford, a combination of skills that rarely co-occurs, and a location restriction that removes most potential candidates.
For two months, no one notices. Applications are low, and the blame falls on the market. The brief stays unchanged until the next hiring round.
At the start, you should state which requirement is causing the most damage, whether the combination is realistic, and what would happen if one of the constraints changed.
Any market comparison must be honest about its own limits, because published ranges are drawn from a mix of markets and levels, and calibration can become incorrect when you apply it in a new direction.
How the agent is built in Agent Studio
The system prompt requires the agent to name the single requirement that most restricts the pool, and to give the source and date for any market claim it makes.
It also needs a "what would change if" section: the location constraint is removed, the role is split, and the level is changed. Calibration without an alternative is only criticism.
Prompt skills hold your own hiring history: which positions took the longest to fill, which job descriptions were revised during hiring, and your actual time-to-fill figures. Your own data is a better comparator than any published range, and most teams never use it.
A Scrappey node and an HTTP Request node collect similar public advertisements. A PostgreSQL node supplies your employment history. Agent Chat then puts the information together.
Notion receives the calibration before the advert is published. That is the only time the calibration is useful.
The agent does not give information about legal requirements. The questions you can ask candidates vary by jurisdiction and should be left to your own advisers. The prompt focuses on feasibility, not on what is permitted.
What this agent is built from
- System prompt: name the single most restrictive requirement, with source and date on any market claim.
- System prompt: a "what would change if" section, so the output is actionable rather than critical.
- Prompt skills: your own hiring history, which beats any published range as a comparator.
- Scrappey (Util): gathers comparable public adverts. HTTP Request (Util) covers other sources.
- PostgreSQL (Data): supplies your real time-to-fill and revised-brief history.
- Notion (Integration): receives the calibration before publication, which is when it matters.
- When an ActionFlow is enough: If you only need to compare a draft against your own past adverts, build the workflow. That comparison catches requirement creep on its own.
Frequently asked questions
Start building AI workflows
Create a free account, open a template or a blank canvas, and run your first ActionFlow.
Newsletter
Get product updates
New nodes, agents, and product notes. We send mail only when we have something worth opening.