AI automation for recruitment: a practical guide
How recruiters automate application triage, interview scheduling, candidate updates, and compliance chasing, and how to build those automations yourself without waiting on a developer.
Amman Vedi10 min read
On this page
Recruitment is a coordination job wearing a judgment job's clothes. The part that needs a recruiter is assessing whether someone can do the work and whether they want it. Almost everything around that assessment is moving information between people: an application arriving, a shortlist going to a hiring manager, three calendars failing to agree, a candidate waiting to hear back, a compliance document that still hasn't been returned.
That surrounding work is structured, repetitive, and remarkably similar from one vacancy to the next, which makes it a strong candidate for automation. This guide covers where recruiters' time actually goes, the automations teams build around it, and how to build them without filing a ticket and waiting a quarter.
What we'll cover
Where recruiters' time actually goes
Research published by Totaljobs in August 2025, based on a survey of 748 HR leaders, put the administrative load at 17.7 hours per vacancy. That's more than two working days of admin for every role filled, and the survey valued it at roughly £17,000 of lost productivity per recruiter per year.
The breakdown is the interesting part, because it maps almost one to one onto things software can carry:
- 3.6 hours reviewing applications
- 2.5 hours scheduling interviews
- 3 hours processing post-interview notes
Those three items alone account for more than nine hours per vacancy, and none of them is the work a recruiter is actually hired for. Nobody's judgment improves because they personally moved a calendar invite. The remaining hours go on the same category of task: chasing documents, updating hiring managers, formatting CVs, re-keying the same candidate details into a second system.
This is a volume problem as much as a time problem. A recruiter carrying eight open roles is carrying eight copies of the same coordination overhead, which is why the load scales badly exactly when hiring picks up.
Eight automations recruiters build
Each one follows the same shape: something happens, information moves, and a person picks up a decision that's ready to be made rather than one that still needs assembling.
1. Application triage
At a glance: applications arrive sorted against the role's actual requirements, with the reasoning attached.
A new application triggers the automation. The CV and application answers are read, checked against the criteria you defined for that role, and sorted into a shortlist, a maybe, and a decline, with a short written rationale for each. The recruiter reviews the sorting rather than doing it.
The rationale matters more than the sorting. A shortlist you can't interrogate is one you can't defend to a hiring manager or correct when it's wrong, so it's worth insisting the output explains itself.
2. Interview scheduling
At a glance: availability is collected, matched, and booked, instead of negotiated over email.
The automation offers the candidate slots that already work for the interview panel, books the one they pick, sends invitations with joining details, and handles the reschedule when someone drops out. This is the 2.5 hours per vacancy in the figures above, and it's the single most commonly automated step in hiring because it's pure coordination with no judgment in it at all.
3. Candidate status updates
At a glance: candidates hear where they stand on a schedule, without a recruiter writing each message.
Going quiet on candidates is the most common complaint about hiring processes and the one most likely to damage an employer's reputation, and it happens for an entirely mundane reason: the recruiter is busy with the next role. An automation that tracks how long each candidate has been at their current stage and drafts an appropriate update keeps that communication steady.
Review before sending: a rejection that goes out with the wrong role name is worse than one sent a day later.
4. Interview notes into structured scorecards
At a glance: the notes an interviewer actually took become a comparable, structured record.
Interviewers write notes in whatever form suits them, which makes candidates hard to compare and makes the three hours per vacancy of note processing necessary in the first place. An automation that takes raw notes and structures them against the role's criteria produces something you can line up side by side, while preserving the interviewer's own observations.
5. CV formatting and client submission
At a glance: candidate profiles come out in your house format, ready to send.
For agencies, reformatting a candidate's CV into the agency template, with the right anonymization and the right emphasis for the client, is a task performed hundreds of times a year with no variation in method. It's a good early automation because the output is easy to check at a glance.
6. Compliance and document chasing
At a glance: outstanding documents are tracked and chased without anyone keeping a mental list.
Right to work, references, certifications, and background checks all arrive at different times from different people. An automation that tracks what's outstanding per candidate and follows up on a schedule removes a task that's pure memory work, and memory is where this one fails.
Keep a person on the verification itself. Chasing a document and deciding whether it's valid are different jobs, and only the first should run on its own.
7. Job ad distribution
At a glance: one approved description goes out everywhere it needs to go, in each board's format.
Posting the same role to several boards, your careers page, and a couple of communities means reformatting the same text repeatedly. Worse, when the description changes, the copies quietly disagree. Publishing from one source keeps them consistent.
8. Placement handover
At a glance: an accepted offer triggers the paperwork and the onboarding handover.
Acceptance is a well-defined moment with a well-defined set of consequences: contract generation, system records, a handover to onboarding, an invoice for agencies. It's a natural automation trigger precisely because nothing about it is ambiguous.
Why good ideas stall before they get built
Ask a recruiter which parts of their week they'd automate and you'll get a precise answer immediately. They know exactly where the time goes, because they've felt every hour of it.
What they typically can't do is build it, and so the idea enters one of three queues.
The developer queue. The request goes to IT, where it competes with the systems migration and the security review. Recruitment operations work rarely wins that competition, and by the time it surfaces the process has often changed.
The specialist queue. A consultant or automation specialist is engaged. This works, and it costs both money and calendar time. It also creates a dependency: when the process changes, the change goes back through the same queue.
The vendor queue. The feature is requested from the applicant tracking system vendor, where it joins a roadmap that isn't yours.
CodeWords is built for the person stuck in those queues. You describe the outcome you want in plain language, and Cody, the automation builder, takes it from there: building the automation, connecting it to the tools you already use, and deploying it. The person who understands the process is the person who builds it, which removes the translation step where most of the detail gets lost.
When the process changes, you describe the change. Recruitment processes change constantly, more than almost anyone outside the function expects. A new client wants CVs in a different format, a hiring manager wants a different screening question, a role in a new jurisdiction needs a different compliance step. A change you can describe in a sentence doesn't need to become a project.
Automations connect to more than 3,000 integrations, which covers the applicant tracking systems, calendars, email, job boards, and spreadsheets most recruitment teams already run on. The free plan covers light use, with Pro at $39 per month and Business at $100 per month as volume grows. Current details are on the pricing page.
One boundary worth drawing deliberately: the automations above prepare decisions rather than making them. Hiring decisions affect people's livelihoods and carry legal obligations in most jurisdictions, and the sensible approach is to let automation handle the assembly and keep a person on every judgment that affects whether someone gets a job.
How to build your first recruitment automation
Start with scheduling. It's the most common first automation in recruitment for good reasons: it happens on every role, it involves no judgment, the time cost is well documented, and success or failure is immediately obvious.
Write down the process as you'd explain it to a new coordinator. What starts it, who needs to be involved, what information is needed, where that information lives, and what the finished result looks like. That description is itself the input, rather than preparation for a separate specification.
Name the exceptions. The candidate who needs an accessibility accommodation. The executive search where nothing is templated. The client who insists on their own forms. Describing what should happen when the automation shouldn't proceed on its own is what makes it safe to rely on.
Run it alongside the manual process for a fortnight. Compare the outputs as you go. Where they differ, describe the correction.
Then take the next one. Application triage is a natural second step, because by then you'll have a clear sense of how precisely you need to describe your criteria.
Frequently asked questions
Does automated screening introduce bias?
It can, and this deserves a real answer rather than reassurance. Automated screening reflects the criteria it's given, so criteria that stand in for something else will carry that through at speed and scale. Two practical safeguards: insist the automation explains each decision so you can inspect its reasoning, and use it to sort and prioritize rather than to reject outright, so a human sees the full pool. Several jurisdictions now regulate automated employment decision tools, so check what applies where you hire.
Do I need an applicant tracking system for this to work?
No. Automations connect to an ATS where you have one, and to a spreadsheet, shared inbox, or form where you don't. Smaller agencies often automate around exactly the tools they already use rather than adopting a new system first.
Will candidates be able to tell it's automated?
They'll notice they're getting timely, accurate updates, which is the opposite of the usual complaint. The things candidates react badly to are generic messages that ignore what they actually applied for and messages sent at inappropriate moments, both of which are content problems rather than automation problems. Review anything going to a candidate before it's sent until you trust the output.
What about candidate data and privacy rules?
Worth answering deliberately rather than generally. Consider where candidate data is processed and stored, which services a given automation touches, what your privacy notices already commit you to, and what your jurisdiction requires for automated processing. Many teams start with automations that touch internal process only, such as scheduling and note structuring, before extending to anything that evaluates a candidate.
Does this replace recruiters?
The pattern teams describe is a change in the mix of work rather than a reduction in headcount. What automates well is the coordination layer: scheduling, chasing, reformatting, re-keying. What remains is sourcing, assessment, negotiation, and advising hiring managers, which is the work recruiters are actually measured on.
How long does it take to build one?
A straightforward automation can be built and deployed in a single sitting. The realistic constraint is usually connecting to the systems involved and agreeing internally on how exceptions should be handled, rather than the build itself.
What should an agency automate before an in-house team does?
CV formatting and client submission, because agencies do it at a volume in-house teams don't, and the output is easy to verify. In-house teams usually get more from scheduling and status updates, where the pain is felt by candidates and hiring managers rather than by the recruiter.
How do I measure whether it's actually working?
Pick the measure before you build, because retrofitting one is how automation projects end up defended on vibes. The strongest measures in recruitment are time to first response, time from application to decision, and the share of candidates who reach a decision rather than going quiet. All three are things candidates feel directly, and all three are countable without a new reporting system.
Time saved is the weakest measure, oddly, because the hours rarely disappear. They move to sourcing and to conversations with hiring managers, which is the outcome you want but harder to point at in a review. If you need a number for a budget conversation, the 17.7 hours per vacancy above is a reasonable baseline to measure your own process against.
Related reading
- AI automation for small business
- How to automate meeting scheduling
- AI automation examples
- CodeWords integrations and templates
Recruiters don't need convincing that the coordination layer is expensive; they pay for it every week.