After the break at Meeting of the Minds 2026, four people took the sofa at Cirkus in Stockholm: Olivier Sibony, Paul Sackett, Kajsa Asplund, PhD – psychologist, author, and Alva board member, and Linnea Bywall, VP People at Quinyx.
Their starting point was new research from Alva Labs and Novus. It surveyed Swedish hiring managers, people working in HR and talent acquisition, and 152 candidates. The questions came from the findings and from the audience.
The host opened with one question: what is the biggest secret to great hiring?
"I think structure is the keyword of keywords. But also, if I had to point something out: making quality of hire your most important KPI. It will shock you how much that will change your behaviour." - Kajsa Asplund
Bywall's answer was shorter:
"I don't look at applications. I look at the structured data."- Linnea Bywall
The first finding: six in ten Swedish hiring managers say they have rejected a candidate who scored well in a structured assessment, a test or a scored interview, because their gut said otherwise.
Sibony pointed back to Sackett's keynote. In theory, we want people to follow what the structured assessment tells them. We also want them to own the decision. There is a tension between the two. The ideal is a hiring manager whose gut agrees with the structured interview, and you get there with the tools Sackett described that make people want to trust the process.
Bywall described what she does when a hiring manager goes against the recommendation. In the moment, she explores it with curiosity: How did you get to that? What is it that you see? How does it link to what we're actually looking for?
Long term, the fix is clear expectations of what the person should deliver in the job. Her team works with competency frameworks, clarified values and leadership principles, so everyone knows what good looks like.
"It's a lot easier to compare structured data to the expected outcome." - Linnea Bywall
Sackett added a test. Is the hiring manager simply rejecting the algorithm, or do they have additional information that is meaningful? He chairs the board of a large theatre in Minneapolis-Saint Paul. During a structured process for a new artistic director, the board learned that one candidate had been allowed to resign quietly from a previous role over misconduct. That was new, powerful information, and a good reason to overrule the scores.
The question to ask is whether everyone else would agree the information is disqualifying. Supporting a different football team than the interviewer doesn't count.
The research also asked 175 managers to pick the top three things that make a hiring process succeed. The personal conversation made the list. Gut feeling topped it. Not one picked quality of hire data.
So what would it take to change that? Asplund's short answer: start measuring it.
Most organisations already have some outcome data, like retention and the reasons people leave. From there, start simple. Ask hiring managers how happy they are with their new hires three and six months in. A short survey is enough to begin with.
Smaller organisations often object that 20 hires a year is too few for a statistical model. That might be true, but even a small sample can get qualitative feedback loops going.
Asplund's shared an example from an Alva customer that had made a clearly wrong hire for a sales role. The new hire struggled to negotiate big deals. The company had a very assertive TA manager, and an AI note taker in every interview. Going back through the notes, they found that the candidate had explicitly said in the interview that he had limited experience negotiating big deals. Nobody had followed up. That led to a valuable conversation about why the information got lost, and what to do differently next time.
Later, the panel was asked about the biggest mistake in the business. Bywall pointed to the distance between the hiring decision and the outcome. With notice periods and ramp-up time, nine months can pass before anyone sees the result, and no one is held accountable for the process.
"Therefore, mediocrity can continue." - Linnea Bywall
Sackett agreed, and said he doesn't understand it. In his research, following up on hires is routine, so we know how to do it. It's just rarely done outside a special study.
Sibony named a paradox. For about a century, the science has been clear about what works in hiring. Yet real people in real organisations keep doing what doesn't work.
One explanation is that the message needs to be louder. After a century of trying, he doubts it. His alternative, which he called slightly politically incorrect:
"We say we are trying to hire the best people, but maybe that's not actually what we're trying to do." - Olivier Sibony
Maybe the goal is finding someone good enough to fill the job right now. Sibony called it satisficing. For some jobs, good enough may be fine. For others, the difference between good enough and truly great is huge. Sibony's advice is to ask which jobs are which.
Sackett took it further when asked about the biggest challenge in the field: persuading leadership to invest in hiring. In many jobs, the difference between a high performer and a low performer is enormous.
"The notion of saying 'I'll just settle for good enough' means your organisation is saying: we stand for mediocrity." - Paul Sackett
Another finding: among TA professionals in Sweden who use AI in hiring, only one in five can say what it was trained on.
Bywall's take: we don't ask enough questions. Hiring is classified as high-risk use of AI under the EU AI Act, and its guidelines work as a checklist. Ask about the training data. Ask whether there is a human in the loop, and whether decisions are logged. Ask who is behind the system. Then check usability: does it solve a real problem for your team?
Asplund added that large language models are hard to inspect, even for experts. Asking what a model was trained on may not get you a good answer. What you can control is what you give the model and how rigorously you do the prep work: what you're looking for, how you'll measure it, the anchors for your interview questions and the weighting of the scores.
"I think to some extent we might over-index on AI and underestimate the power of a pretty simple, elegant structure." - Kajsa Asplund
Sibony separated bias in AI from bias in us that AI reflects. He told the story of Amazon's hiring algorithm, which favoured men because it was trained on Amazon's own past hiring decisions. Amazon shelved it and went back to the old way of hiring, which had created the bias in the first place.
"You're like someone who looks at the mirror and says, 'Oh my God, I don't look good. I have an idea. Let's break the mirror.'" - Olivier Sibony
His point: the algorithm is most likely holding a mirror to you, and that is worth paying attention to.
An audience member asked whether some noise is worth listening to. By definition, Sibony said, noise is unwanted variability. Disagreement is welcome in plenty of places: creative work, matters of taste, markets where one person buys and another sells.
When we're making a judgment we agree has a best answer, noise is a problem. You can still put disagreement to use. Structure the assessments, then ask each person why they rated the candidate the way they did on each one. That gets you to a better answer.
Asplund answered that TA roles will absolutely still exist in five to ten years, but look different. In forward-leaning organisations, much of the admin is already automated. The fast-growing, AI-native companies in Sweden have removed almost all of it and moved that time to face time with candidates: persuading people to join, coaching them to take the leap, and finding the right role for them. The hard part of the job stays.
The panel closed with one tip each.
Olivier Sibony: It's a hard job. You will be wrong from time to time. That's okay. Keep trying to get better.
Kajsa Asplund: Own a hard KPI. Dare to own quality of hire and co-own team performance. It earns you a seat at the table and raises your own motivation.
Paul Sackett: Triage. Nobody can build an elaborate selection system for 143 job titles. Start with your highest-volume roles and the roles where getting it right makes the biggest difference.
Linnea Bywall: Ask every hiring manager whether they actually need the hire. If they say yes, ask why.
That's the gap Alva closes: structure, validated science and outcome tracking built into one system, with a person making every decision.
Next on stage, Malcolm Burenstam Linder, CEO and founder of Alva Labs, asked what the future of hiring looks like when you put all of this into practice. → [Read: Malcolm Burenstam Linder on the future of hiring]