All articles

Recruiting agency automation: search, score, draft, approve

Recruiting agency automation takes the search, the first pass at screening, the outreach drafts and the status updates off a recruiter’s day, so the recruiter spends time on calls and placements. A good setup searches job boards, groups and the web every day, scores each match against the role’s must-haves, drafts a message for you to approve and keeps candidates and clients updated so nobody goes quiet.

I should be upfront about where my experience comes from. I have not published a recruiting agency case study. What I have built is Joven, an AI agent that runs my own remote job search. It does the same work a recruiter does, pointed the other way: it looks for roles that fit one candidate instead of candidates that fit one role. Here is what I learned building it and how it maps to an agency.

What Joven actually does

Joven, my AI job agent, is a Telegram agent running on my own n8n server. Every morning it:

  1. Searches job boards, Facebook groups and the open web for remote roles.
  2. Scores each post for fit against my skills, and for freshness.
  3. Drafts a tailored application for the good ones.
  4. Sends me the drafts in Telegram. One tap sends an approved application as a branded email from my own domain, and every one is archived.

As of 5 November 2026 it had tracked 682 job posts and sent 89 email applications. Every one of those 89 went out because I approved it. None went out on its own.

How the same engine works for a recruiter

Flip Joven around and you have the core of recruiting automation.

  • Search. Instead of one person’s skills, the agent holds the role: must-haves, nice-to-haves, location, rate. It searches the places you already source from, every day.
  • Score. Each match gets a score and, more important, a reason. “Has five years in the stack, no client-facing work mentioned” is more useful than a number.
  • Draft. For strong matches, the agent drafts outreach that mentions what in the profile fits the role. It waits for you.
  • Approve. You read the shortlist, fix or reject drafts, and send the good ones.

The scoring is where the time goes back to you. In my own search, most posts that look right at first glance are stale, wrong for the location or missing something that matters. Having the agent filter those before I see them is the whole point. A recruiter skimming the same profiles for the third time this week has the same problem.

One practical note: each board and platform has its own terms about automated access, and those deserve a careful read before anything is set up to search them. Some sources are better reached through an official feed or API, and some are best left to manual search.

Why a person approves every message

This is the rule I would keep even if you trusted the AI completely.

A recruiter’s reputation lives in their outreach. One wrong-name message or one pitch for a role the candidate is plainly wrong for, sent to a hundred people, does real damage. Drafting is cheap. Sending is the part that carries your name.

I run the same rule across my client channels. My AI agents that watch client chats read WhatsApp, Telegram and Slack, summarise what matters and draft replies, and every outgoing WhatsApp message is a single-use draft that only sends after an explicit approval. I wrote about why in AI agents that draft replies for approval. For a recruiter, it means you review a stack of ready drafts in ten minutes instead of writing them over two hours.

Screening: score and explain, do not decide

Screening is where I would be most careful.

AI is good at reading a CV against a list of must-haves and saying what matches and what is missing. It is not a fair judge of who deserves the job, and it can pick up bias from the way roles and past hires were written. Some cities and states now have rules on automated tools in hiring, so check what applies where you recruit before you let software make any decision about a candidate.

The setup I recommend: the agent scores each applicant against the role, shows the reasons next to the score, and a recruiter decides. Nobody is rejected by a number.

Keeping candidates and clients updated

The other half of recruiting automation is the part nobody enjoys: telling people where they stand.

  • Applicants who never hear back. A short, honest update when a role is filled or paused costs nothing to send automatically and protects your name.
  • Candidates between stages. Interview reminders, prep notes and a check-in after each stage keep good people from drifting to another agency.
  • Client updates. A shortlist summary or a weekly status note drafted from the pipeline, which you check and send.

These run from your ATS if it has an API, or from a CRM pipeline if it does not. They stop when the person replies, so a candidate who answers is talking to you, not to a sequence.

When automation is not worth it

If you place a handful of senior roles a year from a network you already know, an agent searching boards adds little. Your edge is the phone call and the relationship, and no workflow replaces that.

Automation pays off for agencies that fill many similar roles, source from the same places every week, and lose candidates in the gaps between stages. That is volume and repetition, which is what agents are good at.

Getting started

The recruiting and staffing page covers sourcing, screening and follow-up, and the AI agents service page explains how I build agents on Claude and n8n with a person in charge. For the wider set of repeat tasks, see AI automation services.

Send the project brief at /#brief with the roles you fill most, where you source today and which part of the week eats your time. I reply with a written plan and the price, and we can talk it through on WhatsApp after that.

Questions

What can a recruiting agency automate with AI?

Searching boards and the web for matches, scoring each one against the role, drafting outreach, sending interview reminders and writing client status updates from the pipeline. The decisions about who to put forward stay with the recruiter.

Should AI decide which candidates get rejected?

No. Let it score and explain why, and let a recruiter make the call. Some places now have rules on automated hiring tools, so check what applies where you recruit before you automate any decision.

Will the AI message candidates on its own?

Only if you decide it should. By default every outreach message is drafted and waits for your approval, which is how I run my own agent.

Does recruiting automation work with my ATS?

If your ATS has an API, yes. If it does not, candidates and stages can live in a CRM instead, which is a bigger change and worth thinking through first.