You don't need AI everywhere to hire better

The question is not: where can we use AI?

The better question is: where can technology help us have a better conversation?

Many recruiters are caught between two concerns. On one hand, the pressure to do something with AI keeps growing. On the other, there is uncertainty about privacy, regulation and how much automation candidates will still experience as human.

That is no reason to stand still. It is a good reason to start small.

The European AI Act follows a risk-based approach. AI systems used for recruitment and selection can fall under the rules for high-risk AI, depending on their use and classification. The GDPR also sets limits on decisions based solely on automated processing when those decisions produce legal or similarly significant effects for people. Neither framework says that you can never use AI. Both do make clear that purpose, context, transparency, human control and accountability matter.

This is general information, not legal advice. The concrete obligations depend on the application, the organisation’s role and how the system is deployed.

Start with one bounded problem

Using AI because it is on the roadmap is not a strategy. A small experiment with a clear purpose can be.

Consider a use case that:

• Prepares a human conversation: technology helps surface relevant observations or follow-up questions, but does not take over the conversation.
• Adds behaviour instead of pretending to predict it: you gather more information about how someone makes choices, reasons or responds to new information, without pretending that a model can predict a person’s future.
• Stays bounded and controllable: recruiters know what the system does, which data is used and where their own judgement remains essential.

This kind of boundary makes it easier to ask the questions every responsible AI project should answer: what exactly are we trying to improve? Which data is necessary? What happens when the output is wrong? And can a candidate understand what role the technology plays?

No AI theatre. No automatic hiring manager.

The other extreme is just as risky: a system gives a candidate a score and quietly steers the decision that follows.

A recruiter does not automatically become a better decision-maker because a dashboard shows a number. And a candidate does not become less human because technology supports part of the process.

The value often lies in the space between those two extremes. Technology can help collect information and make patterns discussable. People are still needed to understand context, ask questions, spot exceptions and take responsibility for a decision.

Human oversight also means more than clicking a button after the fact. A person must be able to understand what is happening, critically assess the output and intervene when the application is unsuitable or unreliable.

Observe choices, not just stories

A CV and an interview tell us something important, but they mainly show how someone presents themselves in a familiar selection setting. An additional perspective can make the conversation richer.

At Assess4me, we use game-based assessments to make behaviour more observable. In a bounded game environment, recruiters can see for example how someone:

• Makes choices: what gets priority when not everything can happen at once?
• Approaches problems: do they try something quickly, analyse first, or change direction after feedback?
• Adapts: what happens when the situation changes and the first solution no longer works?

That is not an automated verdict on who the best candidate is. It is an additional source of behaviour-based information that a recruiter can use to ask better questions and have a more nuanced conversation.

Starting small is not a lack of ambition

A good first AI application in recruitment does not have to automate the entire process. It should make clear which problem you are solving, which boundaries you respect and where people keep the final say.

Starting this way builds more than technology. It also builds trust with recruiters and candidates.

Because better selection does not happen when AI replaces people. It happens when technology helps us see more of human behaviour, while people remain responsible for handling that insight with care.