AI can already help write job descriptions, organise applications, screen for clear requirements, schedule interviews and summarise information. I see real value in that.

But speed is not the same as quality. We still see roles stay open for a long time, disappear, and then return with almost the same job description. That tells me that many hiring problems are not simply screening problems. They can start with an unclear role, unrealistic requirements, slow decisions or a process that was not well designed in the first place.

AI can reduce the noise

Anyone who has worked in recruitment knows what a busy application pipeline can look like. Some applications are clearly not relevant to the role. A person may not have a technical skill that is genuinely essential, the required work authorisation, a language that the job truly depends on, or another clear requirement.

This is where AI can be useful. It can reduce the first layer of noise and turn a very large pipeline into something a recruiter can work with.

But the criteria matter. They should be limited, clear and genuinely necessary for the job.

A degree level, a specific number of years of experience or a familiar job title should not automatically become a rejection rule just because they appeared in an old job description. A recruiter still needs to ask what is really essential, what is only preferred and what the market can realistically provide.

The recruiter should not become a system operator

I do not think the future value of a recruiter is in reading hundreds of CVs one by one. It is in understanding the market, challenging unclear requirements, helping the hiring manager define what matters and making better decisions about talent.

If AI removes repetitive work, that should create more time for the parts of recruitment where judgment adds value.

That also means recruiters need to question the process itself. Are we asking for the right things? Are the requirements realistic? Could we be excluding good people for the wrong reasons?

What happens to the candidates we never see?

Imagine 1,000 people apply and a system shows the recruiter the 30 strongest matches. The recruiter can review those 30 carefully. But how do we know there were not another five strong candidates among the people who never reached that stage?

Maybe their job title was different. Maybe their career path was less traditional. Maybe they had transferable skills described in a different way. Or maybe they simply did not know how to write a CV that the system would recognise.

Candidates are already being told to rewrite their CVs around keywords so that applicant-tracking systems and AI tools will notice them. I find that uncomfortable. If AI is really making recruitment smarter, it should eventually help us move away from the keyword game, not make people better at playing it.

If the company cannot explain a rejection, should it trust it?

This is one of the questions I keep coming back to.

If an automated system has a meaningful role in screening or rejecting someone, the company should at least be able to understand why. The explanation does not need to be long or technical. It might be that a required work authorisation was missing, an essential language level was not met, or a genuinely critical technical skill could not be found.

If the company itself cannot understand why the system rejected someone, I would question how much it should trust that rejection.

For me, this is not only about candidate communication. It is also about the quality of the company’s own decision. If a decision cannot be explained internally, it becomes difficult to check whether it was reasonable.

Candidate experience should matter too

Companies often talk about AI as a way to make recruitment more efficient. But efficient for whom?

As a candidate, I have experienced processes where I uploaded a CV and then had to enter the same work history, education and other information again by hand. If AI is meant to make hiring more efficient, I think that efficiency should also be visible to the person applying.

A smarter process should reduce duplication, communicate clearly and help candidates understand where they stand. It should not simply move more administrative work onto them.

I have been thinking about this problem for a while

This is also a question I explored through Veenro, an early-stage AI-supported job-discovery platform we built but never fully launched. The product was coded and reached an early working stage, but we did not continue developing it into a full service.

The idea was to let people describe their skills, interests, language needs and preferences more naturally, instead of repeatedly rewriting themselves for each application. I was interested in a system that could have a more useful conversation with the person, understand more than a job title, and then surface relevant opportunities.

I still think that direction is interesting. A system could show where a role matches someone, where the gaps are and what the person should know before applying. But I would rather let the person make the final decision about whether to apply than have the system decide, “This job is not for you.”

Where I currently draw the line

I am not arguing that AI should stay out of recruitment. I think it can remove a lot of unnecessary work.

But today, I would use three principles:

• The criteria used for screening should be clear and genuinely related to the job.
• Important rejection decisions should still have human oversight.
• The process should create value for the candidate as well as the company.

And I would add one more:

AI can make hiring faster. The more interesting challenge is whether we can use it to make hiring clearer, more useful and better for everyone involved.

The work

Across technology, startups, consultancy, manufacturing and project environments, I have worked on full-cycle recruitment and targeted sourcing for technical, data, business and leadership roles. This included building pipelines from scratch, Boolean search, direct outreach, candidate assessment and negotiation.

Examples include building a small Data Science / Data Analytics team for an international AI company, recruiting a senior DevOps leadership hire, sourcing experienced C++ and mobile professionals for a game-company assignment, and recruiting the initial technical and management workforce during an industrial ramp-up.

What matters to me

A hiring manager may start with a title, a wish list and a sense of urgency. The recruiter’s job is not simply to “take the order.” I like to clarify what problem the hire should solve, which requirements are essential, which are preferences, and what the market is likely to give us.

Source basis: verified CV materials used for recent applications.

Starting point

There was no single neat framework to inherit. The work involved sitting with senior managers and department heads, understanding roles, drafting job descriptions and KPIs, and bringing material into discussions so that expectations could be refined rather than invented in isolation.

How I worked

I prepared drafts before meetings, used the conversations to clarify responsibilities and expectations, documented decisions, and followed up so that the output did not disappear after the meeting.

The problem

UniBuddy already existed, but parts of the experience were loosely structured. I gathered participant feedback, used AI as one tool for organising and analysing recurring themes, identified gaps, and helped create clearer guidance around roles, expectations and participation.

The wider lesson

International employee experience is not only about giving information to the newcomer. It is also about the team: language practices, access to informal knowledge, expectations, and whether the environment makes participation possible.

Public context: University of Helsinki article ↗

My involvement

I contributed across ideas, pilot discussions, stakeholder outreach, potential partnerships and user feedback. I also brought an HR and Talent perspective into conversations about how people might use the product when making study or career decisions.

A practical problem I noticed

Career counsellors were finding parts of the assessment results difficult to explain and use with clients. I created a practical visual guide that broke down models including career interests, work values, work styles, the Big Five and learning preferences into clearer language, examples and counselling questions.

Public context: NexPath LinkedIn activity ↗

What we built

We developed an early-stage AI-supported job-discovery platform where people could describe what they wanted and build a profile around their skills, interests, language needs and preferences. The system was designed to surface more relevant opportunities without relying only on a traditional CV.

The product was coded and reached an early working stage, but we did not continue developing it into a fully launched service.

What I still find interesting about the idea

I remain interested in a more conversational application experience: a system that can learn about a person, show where a role matches or does not match, and give useful information without taking the final decision away from the person.

My role

I originated the project and lead concept, conversation and editorial direction. The work includes guest research, interviews, story structure, coordination with a small creative team and attention to how people are represented.

Professional relevance

The project uses research, interviewing, stakeholder coordination, editorial judgment and cross-cultural communication. Those skills also appear in my professional work with international employees, candidates and partners.

Visit Locker Room Talks ↗