By Matt Kielbasa, Founder of Inflowave (LinkedIn)
Most recruiting teams adopt automation backwards. They start with the part of the process that feels most tedious to them, usually outreach, and end up automating the exact moment a candidate decides whether your company is worth their time. Meanwhile the genuinely mechanical work, the scheduling and the status updates and the pipeline hygiene, stays manual because nobody thinks of it as a project.
The useful question is not “what can we automate?” It is “where does a human presence change the outcome?” Everything else is overhead.
The test: does a person change the result here?
Run every step of your process through one filter. If a candidate would behave differently knowing a human was involved, keep the human. If they would not notice or would not care, automate it.
A rejection email is a good example. Candidates notice a templated rejection, and they notice silence even more, but almost nobody expects a personal note after a first-round screen. Automate the sending, keep the writing human, and write it once properly rather than badly a hundred times.
Contrast that with a final-stage negotiation. The moment a candidate senses they are talking to a workflow rather than a person with authority, your close rate drops. No efficiency gain covers that.
Automate: scheduling
Interview scheduling is the clearest win in recruiting and the one teams delay longest. It is pure coordination, it involves no judgement, and every hour it sits in someone’s inbox is an hour a candidate spends talking to your competitor.
The data supports moving fast here. SHRM’s benchmarking puts average time-to-fill at roughly 44 days, and the single largest recoverable chunk of that is the dead time between stages rather than the interviews themselves. Self-service booking removes most of it.
What to keep human: the confirmation for final-round and executive interviews. A short personal note before someone gives up half a day is cheap and it visibly changes how prepared they arrive.
Automate: pipeline hygiene and status updates
Candidates rank “not knowing where I stand” as one of the worst parts of applying. Recruiters rank “updating the system” as one of the worst parts of recruiting. These are the same problem and automation solves both at once.
Stage changes, reminders, nudges when a candidate has sat untouched for a week, handoffs between recruiter and hiring manager: all of it should fire on its own. This is where a proper pipeline earns its keep, and it is the piece most teams try to run out of a spreadsheet for far too long. A CRM for marketing agencies and a recruiting pipeline solve a near-identical problem, moving people through defined stages without anyone falling out of the middle, which is why the tooling looks so similar.
What to keep human: the reason behind a rejection at final stage. Automate the notification, not the explanation.
Automate: sourcing and enrichment, carefully
Finding candidates, deduplicating them, enriching records, and ranking them against a brief is legitimate automation. It is pattern matching at volume, which software does better than people do at four in the afternoon.
Two guardrails. First, automated ranking should shortlist, never reject. A model that silently discards applicants will encode whatever bias sits in your historical hiring data, and you will not find out for a year. Second, audit the shortlist against the full applicant pool monthly. If the two populations look meaningfully different in ways unrelated to skill, your filter is broken.
Keep human: the first real conversation
This is the step most often automated and it is almost always a mistake.
The first conversation is where a candidate forms their view of your company, and where you learn the things a CV cannot tell you: whether they are actually looking or just curious, what would make them say no, whether the role as written matches the role as lived. None of that survives a form.
AI voice screening is improving quickly and it has a genuine place at very high volume, where the alternative is not a phone call but nothing at all. At normal volumes, a fifteen-minute call from a recruiter who has read the CV beats any amount of automated screening, and it costs less than the tooling.
Keep human: anything involving a judgement call about a person
Salary flexibility. Whether a non-obvious background is worth a punt. Whether to keep a strong candidate warm for a role that does not exist yet. Whether the hiring manager’s stated requirement is real or aspirational.
These are the decisions that determine whether you hire well, and they are exactly the decisions that look automatable on a process diagram because they sit at defined stages. Resist it.
The sequence that actually works
Teams that get value from recruiting automation tend to follow the same order.
Start with scheduling, because it is self-contained, the return is immediate, and it does not touch candidate judgement. Then move to status updates and pipeline automation, which is where you stop losing people to neglect. Then sourcing and enrichment, with the shortlist-never-reject rule in place. Only after those three are stable should you look at anything that talks to a candidate unsupervised, and even then only where the volume genuinely exceeds what humans can cover.
Teams that fail usually invert this. They buy a tool that promises end-to-end automation, switch it on across the whole funnel, and discover six weeks later that their offer-accept rate has slipped and nobody can say exactly why.
Picking tools without over-buying
Most recruiting teams need far less software than vendors suggest. A pipeline that reliably moves candidates through stages, scheduling that candidates can self-serve, and automated reminders will cover the large majority of the value. The comparison thinking is the same as choosing any marketing agency software stack: buy for the workflow you actually run, not the one in the demo.
Two questions worth asking any vendor. Where does a human have to intervene for this to work, and what happens when they do not? And what does this look like when the pipeline is empty, which is most of the time for most roles?
The short version
Automate coordination. Keep judgement. The distinction is not about how advanced the technology is, it is about whether the candidate’s decision changes when a person is present.
Teams that hold that line get the efficiency without the quiet cost. Teams that do not end up with a very fast process that fills fewer roles.
