Automation can locate people faster, organise applications, suggest matches and help recruiters write first drafts of outreach.

Moreover, this is useful. However, in recruitment automation, if a tool screens hundreds of profiles quickly, when should a person step in? The answer depends on judgement and context. The answer is that attention varies during the search, especially when a hiring decision forms. Throughout the search, attention changes at each stage. The responsibility for defining, testing and interpreting the result does not disappear.

Start with the work, not the job title

Imagine hiring a Product Manager. A search tool can find hundreds of people with that title. However, “Product Manager” or “Product Owner” can mean very different things across companies. One person may own discovery, commercial outcomes and a roadmap. Another may mainly manage a delivery backlog. Someone with a different title may have exactly the experience the team needs.

To start, the recruiter and hiring manager must agree on the real criteria before asking a tool to select candidates.

• What problem will this person solve in the first six to twelve months?
• Which outcomes will they own, and how will success be measured?
• Which experiences are essential on day one? Which can be learned?
• What evidence would demonstrate the required experience, regardless of job title?
• What does this team need from a new colleague in the way they communicate, decide and collaborate?

Additionally, use the shortlist as a hypothesis, not a verdict

This is where human judgement first shapes the automation. If the criteria are vague, a faster search will produce a larger set of results. Automation in recruitment complements automation by helping identify matching profiles. Automation is valuable for finding possible matches, organising information and prioritising a review.

If the shortlist feels wrong, look at some of the profiles the tool left out. Additionally, are strong candidates using a different title? Moreover, is relevant work described under an unexpected industry? Furthermore, did the search favour exact keywords over evidence of ownership or results? Finally, are the requirements too narrow?

Then adjust the criteria and search again. This is a feedback loop: define, search, inspect, refine. The human does not need to read every CV in the same depth. However, they should inspect enough selected and excluded profiles to understand what the tool is finding and missing.

Make outreach personal where it matters

AI can help structure a message, summarise a public profile or prepare a first draft. It can save time on the blank page. Yet a candidate can usually tell when the message could have been sent to anyone.

A useful outreach message should explain why this person was approached. It should state what work the opportunity involves. It should explain why a conversation might be relevant to their experience or ambitions. Additionally, this process supports automation in recruitment.

Automation can assist with preparation and administration where useful. Afterwards, review the claim, tone and reason for contacting that particular person. Personalisation is more than inserting a first name or company name.

Change the level of human involvement as the process develops

At the beginning, technology can help with sourcing, sorting, scheduling and recording consistent information. As the process moves toward assessment, the recruiter and hiring manager need to take a more active role.

An interview is where we test the evidence behind a profile. Additionally, what did the candidate personally own? Moreover, which decisions did they make? Furthermore, what trade-offs did they face? Finally, what changed because of their work? In addition, consistent questions and a shared scorecard help interviewers compare relevant evidence rather than rely on confidence or similarity.

It also notes which business goals this hire must support.

Team context also matters. I would say matters a lot. A hiring manager knows why the role exists, where collaboration has been difficult. That information needs to become clear, job-related assessment criteria.

“Team fit” on its own is too vague to be a useful decision. Ask instead how someone works across functions, handles disagreement, communicates decisions or operates in the conditions this team faces. Assessment works in both directions.

Candidates need a candid view of the role, manager, team and expectations. A recruiter can align what the candidate wants with what the organisation offers. That conversation cannot be reduced to matching keywords.

This final decision is a conversation, not a score.

In the end, the team should explain why the chosen candidate can do this job using evidence from the process. The recruiter can surface gaps between interviewers, verify if criteria changed midway, and help decide what still needs verification.

Next is another human task: helping the candidate decide. An offer is more convincing when the team can speak honestly about the work and growth opportunity. The manager and the challenges ahead also matter.

A practical division of work

Stage	Let automation help with	Keep people accountable for
Role definition	Drafting a scorecard and organising inputs	Agreeing on outcomes, evidence and essential criteria
Search	Finding and grouping possible candidates	Testing search logic and reviewing missed profiles
Outreach	Research preparation, drafts and follow-up reminders	Relevance, accuracy, tone and a genuine reason to connect
Assessment	Scheduling, notes and consistent question sets	Follow-up questions, evidence, context and fair comparison
Decision and offer	Summarising agreed evidence and tracking steps	The hiring decision, candidate conversation and closing

There is no single point at which human judgement suddenly begins.

Additionally, People need to design the search, check its results, assess the evidence, and build trust on both sides.

We should decide which part to automate to gain time for conversations and decisions that ensure a successful hire.