This is part of our ongoing series on how we’re building the agentic enterprise at ConductorOne. In previous posts, we’ve explored what happens when AI agents become the primary execution layer of knowledge work. This post is about what that looks like in practice – inside our own recruiting function.
An ATS. A sourcing tool. A scheduling tool. Scorecards. Notes. Reporting. A Slack integration. Five Zapier automations nobody remembers building.
Every year the stack gets bigger, and somehow the work never gets easier.
So we stopped buying tools and started building them.
The Stack We Were Embarrassed to Admit We Had
Here’s the honest version of a “modern” recruiting stack:
| Function | Tool |
|---|---|
| ATS | Ashby / Greenhouse |
| Scheduling | GoodTime / Calendly |
| Sourcing | LinkedIn + JuiceBox |
| Scorecards | ATS |
| Candidate notes | Also the ATS |
| Reporting | BI tool nobody checks |
| Outreach | Outreach / JuiceBox / Gem |
| Workflow automation | Zapier (good luck) |
Six tools. Six contracts. Six logins. Eight UIs your team has to context-switch between – just to move one candidate through one process.
And every year, a vendor pitches you a seventh.
Recruiting is one continuous workflow. Splitting it across eight products isn’t a system. It’s a hostage situation.
The New Stack: Three Components
1 – Slack Is the Interface
Recruiting happens where people already work. So we made everything run inside Slack. Managers can request a new role, submit interview feedback, approve offers, or check pipeline health without ever opening the ATS.
/candidate-feedback
/request-headcount
/interview-score
/pipeline-status
2 – The ATS Becomes a Database
The ATS is no longer the system of interaction. It’s the system of record. AI agents write to it automatically. Nobody logs in. Nobody clicks through stages. It quietly stores the data while real work happens elsewhere.
3 – AI Agents Do the Work Nobody Should Be Doing
Interview Feedback Agent. Asks interviewers how the candidate did – in plain English, in Slack, right after the interview. No forms. No chasing. The agent extracts signals, fills scorecards, and surfaces red flags. Done.
Scheduling Agent. Watches candidate stages, proposes interview times automatically, and reschedules when calendars change. Recruiters are removed from the loop entirely.
Candidate Intelligence Agent. Every hiring manager gets a candidate briefing before the interview – background, concerns, fit against the role – generated automatically. No more “I didn’t have time to review the resume.”
The whole system cost almost nothing to build. No vendors. No implementation consultants. No six-month integration project with a $40k SOW attached to it. Just AI coding tools, APIs, and a small amount of glue code.
What Recruiting Actually Is (And What We Forgot)
Somewhere along the way, the recruiting industry started confusing the job with the software.
We started acting like recruiting is: sourcing candidates, sending outreach sequences, clicking stage buttons, scheduling interviews, filling scorecards. That if you do those things efficiently enough, hiring works.
It doesn’t. Because none of that is the real job. Those are administrative side effects of the real job.
The real value of recruiting has always been talking to people. Understanding what motivates them, what scares them, whether this company is actually the right next step – and being honest enough to tell them when it isn’t. It’s being the voice of the company to candidates, and the voice of the market back to hiring managers who think they can hire a senior engineer for $90k because “we have equity.”
The best recruiters don’t spend their time in software. They spend it:
- Helping a manager understand what they actually need
- Telling a team their comp is uncompetitive
- Saving a candidate from taking the wrong offer
- Closing someone who had three other offers
- Aligning five interviewers who all want different things
Modern recruiting software accidentally inverted all of this. Tools that were supposed to free up recruiter time slowly became the thing recruiters spend all their time on. We optimized the side effects and neglected the actual job.
70% of Recruiting Ops Work Should Not Exist
Let’s say the uncomfortable part out loud.
A massive portion of recruiting operations work – the scheduling, the chasing, the stage updating, the report generating, the nudging, the data copying – only exists because your systems are broken.
It’s not strategic. It’s not valuable. It’s glue work. It exists to compensate for tools that don’t talk to each other, workflows that don’t automate themselves, and processes designed around software limitations instead of human ones.
We’ve told ourselves “this is just part of recruiting” for so long we stopped questioning it. It’s not part of recruiting. It’s part of operating a bad stack.
When you replace this work with AI agents, something counterintuitive happens: the team doesn’t get smaller. It gets sharper.
Interviews get scheduled instantly. Feedback arrives without anyone chasing it. Candidates stop falling through the cracks. Hiring managers stop going dark.
And recruiters? They get their time back. They stop being workflow administrators and start being talent advisors again. Which is what they were hired to be.
The Org That Comes Next
Here’s the uncomfortable implication nobody is saying clearly enough: you don’t need a large recruiting ops function to run a great recruiting org anymore.
| Old Model | New Model |
|---|---|
| Recruiters | Recruiters |
| Recruiting Ops | Systems thinkers / Designers |
| Sourcers | AI agents |
| Tool managers | – |
| Coordinators | – |
This isn’t about cutting headcount. It’s about eliminating work that shouldn’t exist so the people you have can do work that actually matters. Better support for hiring managers. Better experience for candidates. Better hires, full stop.
Two Paths Forward
Recruiting teams are splitting into two groups: those who keep buying tools, and those who start building systems.
The first group’s operational burden will compound every year. The second will run leaner, move faster, and do it better.
The next generation of recruiting stacks won’t be built by vendors. They’ll be built by the teams themselves – in weeks, not quarters, for a fraction of the cost.
The technology to do this exists right now. The question is whether your team is willing to stop asking “which tool should we buy” and start asking “what should we build.”
If your recruiters are spending most of their time coordinating, updating, chasing, and reporting – your system failed them.
Fix the system.
