By: Jay Kt
My worst year in this job was the one where the req board went from nineteen roles to two overnight.
Hiring freeze, in-house, mid quarter, and the instruction was to keep the pipeline warm. So I spent four months doing full cycle work with no cycle at the end of it. Intake calls about roles that might come back. Outreach with no offer behind it. Screening notes filed into Greenhouse against a req that had closed before I finished typing them. It taught me more about the shape of the work than any year of hitting target ever did, because with the closing taken away you can see precisely how much of the job is relay.
Recruiters who run full cycle have a particular relationship with the phrase manual work. They are not defending a stage. They carry the whole chain, from the intake call where a hiring manager describes a role in the vaguest available terms through to the Friday evening a counter offer lands and somebody has to talk it through.
Nobody hands a full cycle recruiter a slate. You build it, work it, close it, and do the coordination in between. Which is why the question of where machines help gets a different answer from a full cycle desk than from anyone who owns one slice of the process. You can see the whole length of the thing. You know where it bunches.
The job is eight modes, not eight tasks
A specialist sourcer spends the day in one mode. Full cycle means switching between a research mode that wants quiet and depth, a persuasion mode that wants energy, an administrative mode that wants nothing but accuracy, and a diplomatic mode for the hiring manager who has just rejected a candidate they approved in writing last week.
The switching is the cost that never appears in a process diagram. Ten minutes of research does not produce a tenth of an hour of research value, because the first four go on remembering where you were. Three live reqs, three technical vocabularies, and by four in the afternoon I have X-rayed the wrong market for the wrong role at least once.
Any honest account of where automation helps has to start with the modes, because that is where the friction actually lives.
Re-typing is the tell
Watch what happens to information as it travels down your own chain. A hiring manager says something specific on intake. You write it in a notebook. You turn it into a boolean string. The results become a spreadsheet. The spreadsheet becomes outreach copy. The replies become ATS records. Then somebody writes a summary for the hiring manager containing roughly what the hiring manager said in the first place.
Every one of those moves re-enters the same information by hand and loses a little detail doing it. The re-typing is the signal. Where a process forces you to restate something you already captured, there is manual work with no judgment in it.
|
Seam |
Re-entered by hand |
Actually a judgment call |
|---|---|---|
|
Intake to search |
The brief, turned into titles and keywords |
Which stated requirements are real |
|
Search to outreach |
Names, companies, the hook for the first line |
Whether this person is plausibly movable |
|
Outreach to screen |
Reply content, availability, money signals |
Whether the reply means yes or means polite |
|
Screen to submission |
Notes, evidence, the reason for advancing |
The recommendation itself |
The case for an AI recruiting platform sitting under a full cycle desk is that it absorbs the left column and leaves the right one exactly where it is. Anything claiming the right column as well is a thing you will spend next spring unwinding.
Intake is the stage worth protecting
Intake is where the req is won or lost and it is almost entirely a human skill. “Senior” means “a person who will not need me in standup”. “Must have experience in our industry” means “the last hire took nine months to understand our customers”. The title on the requisition does not exist outside their own org chart.
Getting to the version behind the version takes follow up questions, a little pushback, and occasionally the nerve to say the role as written does not exist in this market at that money. I have run an intake where the hiring manager wanted a Stripe engineer but cheaper, said exactly that out loud, and could not see why it was a problem rather than a budget line. We lost six weeks before he could.
What can come off your hands is what happens immediately after. Turning that conversation into a search brief is mechanical labour dressed as strategy: plain sentences converted into title variants, adjacent titles, company lists and skill proxies, done from memory, which means the brief is only as good as whether you have filled this role before. Describing the role in plain language and having the brief built out of that description removes a step where nothing is gained by hand except tiredness.
Search eats the middle of the week
The search layer is where full cycle recruiters lose the most time and notice it least, because it does not feel like loss. It feels like working.
Open LinkedIn. Run the string. Read profiles. Adjust. Read more. Somewhere in there a boundary gets accepted that was never a real boundary: the population you are searching is the population that maintains a profile on one network, and in narrow fields the strongest people stopped bothering years ago.
Proof that a candidate can do the work usually sits outside any profile. Patents with their name on them. A long technical answer on Stack Overflow that my own engineers quoted at me before I had ever sourced the author. Courseware written for a certification body. A standards committee seat, or the organiser slot on a meetup that has run for six years. A Slack or Discord community where the person answers other people’s questions for free, which tells you more about how they will behave in code review than any screening call will. Clinical publications on PubMed, if the req is healthcare. Reported decisions, if it is legal.
Also, nobody reads the GitHub. I say that as a sourcer who put repository links in slate notes for two years and got asked a question about one of them exactly once.
Reading across all of that in parallel is the clearest case for handing search to a machine, and the reason is throughput rather than intelligence. One maintainer list can be read attentively. Forty of them cannot, not before lunch, not while running two intakes and rescheduling a panel the hiring manager has already moved twice.
The screening note nobody writes
The fastest decaying thing on a full cycle desk is the record of your own judgment. You screen somebody, form a clear view inside twenty minutes, advance or pass, and then type three words into a field because the next call starts in four minutes. Four months later the same person surfaces on a different req and you cannot reconstruct why you passed. So they get screened again, or skipped to be safe, which is worse.
Across a year of reqs your database becomes a list of names with no memory attached to any of them. The re-engagement that should be your cheapest source of candidates, the person who came second on a role that closed in March and would walk into the one that opened yesterday, quietly stops being available.
A shortlist with no evidence attached is just a list of names, and a screening note with no evidence in it is the same failure one stage earlier. The useful record says what was verified and how, what got asked, what the person said about timing and money, and what would have to change for them to move. Somebody reading that six months on can act on it. Somebody reading “strong, not now” cannot.
Silver medallists and the shelf life of a pipeline
Every full cycle recruiter is sitting on a pipeline that is quietly expiring. People change roles. Numbers stop working. The strong candidate who was mid process when the freeze landed is eighteen months into the job they took instead, and I know that because I went back to the pool I had kept warm through those four months and most of it had moved on without telling me.
Contact data rots faster than anything else in the record, which makes it the most obvious thing to have checked continuously rather than on the afternoon it is urgent. A slate you cannot reach is a document. The candidate who was perfect and unreachable is a specific flavour of frustration, and the failure almost always sat in the contact record, one step downstream of a search that had worked perfectly well.
Refreshing a pipeline never wins an argument for priority, because it produces nothing visible on the day it happens. It shows up later.
Where the line sits at offer
Nothing about a recruiting automation platform helps with the conversation where somebody is deciding whether to disappoint their current manager. That call is yours. So is reading a counter accurately, and knowing when “I need to think about it” means the decision is already made and the candidate is managing you.
What changes on a full cycle desk is the ratio. Less of the week on restating information already captured, more of it on the three or four conversations that decide whether the req closes.
It shows up in how your week feels long before it shows up in a report. You stop being the relay between stages and go back to being the person who reads people accurately, which is the part of the job nobody was ever going to hand over, and the part most full cycle recruiters got into it for.
