AI sees in black and white, people live in the grey

What we learnt at UNLEASH Paris 2026 

In October, the Clevry team spent two days at UNLEASH Paris. Unsurprisingly, AI was everywhere, from the main stage to the coffee queue. But the conversation has moved on. Organisations are no longer asking whether to use AI. They’re asking how to make it work, and what it means for their people.

One idea stayed with us all week. In a session from Google, AI was described as fundamentally binary: yes or no, match or no match, right or wrong. People operate in the grey. They read context, weigh trade-offs, change their minds and make judgement calls that rarely come down to a simple yes or no.

That grey area is where most important talent decisions happen. Who’s ready for a bigger role? Who has the judgement to question an answer that sounds right? AI can’t untangle that on its own, and this is where people science comes in: it makes the grey visible and measurable.

Here’s what else we took away.

1. Work is being redesigned around outcomes

The most ambitious organisations are going further than doing the same things faster with AI. They start again from the outcome they need. Sharon Doherty at Lloyds Banking Group calls this “zero basing”: ignore how the work is done today, define the outcome, then decide which tasks belong to people and which to AI.

SAP has taken this approach in recruitment. Gina Vargiu-Breuer, its Chief People and Culture Officer, described a rebuilt process in which AI agents draft job ads, match candidates and schedule interviews. SAP reports a 95% reduction in recruiter time as a result, and that time now goes into conversations with candidates.

When roles change this fast, though, the frameworks that define what good performance looks like in them go out of date quickly. Siemens, for example, found that 850 job requisitions for a single service engineer role contained 550 unique versions. If your job analyses and competency frameworks don’t keep up, you risk hiring and developing people against a role that no longer exists.

2. AI can take on tasks, but not accountability

Gina Vargiu-Breuer drew a useful distinction between humans being in the loop and on the loop. In the loop means a person approves an agent’s action before it happens, which is essential when the cost of getting it wrong is high. On the loop means people set the guardrails and define the playing field.

Her rule of thumb was clear. Anything that affects someone’s career needs a human to make the final call. As she put it, you can delegate work to agents, but never accountability. And with regulation such as the EU AI Act in mind, being able to explain a decision (and the evidence behind it) is more important than ever.

3. Judgement is the skill of the moment

Asked which human capability becomes more valuable as AI develops, Gina gave a one-word answer: judgement.

She wasn’t alone. Siemens now treats “foundational skills” (cognitive skills, self-regulation and working together) as being just as important as technical ones. Research from The Conference Board pointed to judgement, critical thinking, adaptability and systems thinking. Our own Dr Alan Redman, presenting with Claudia Nuttgens from our partner AMS, set out four skills for working well alongside AI: critical thinking, communication, emotional management and ethical judgement.

These are the skills that live in the grey. They’re hard to spot on a CV and easy to miss in an unstructured interview, but validated psychometrics, well-designed assessment exercises and structured interviews can measure them much more reliably.

4. AI is only as fair as the data behind it

Nigar Huseynova, Group Head of Learning and Development at PASHA Holdings, raised one of the most important warnings of the week: skills data carries historical bias. People who are more visible, or who have worked on bigger projects, leave a bigger data footprint, so AI may rate them more highly. Her answer was to validate the data with assessment before trusting it.

Siemens made a similar point. Context is AI’s weakest spot, and subject matter experts are needed to check whether its outputs make sense. This goes well beyond skills data. Any AI output is only as good as the information it’s based on, whether that comes from an organisation’s own systems or from content online, so it needs checking before anyone acts on it. For high-stakes decisions such as hiring and promotion, AI inferences aren’t enough; you need validated, fair evidence.

5. Adoption is a people challenge

The Conference Board’s research found that AI use is outpacing training, and that training on its own isn’t the answer. The barriers people face are practical: courses that are too basic, a lack of access or permission, no time to learn, and managers who don’t use AI themselves.

The advice across sessions was to start small. Pick a problem that matters, pilot a solution and learn from it. Marci Paino from Cisco summed it up as “progress over perfection”. This fits the behaviour change model Alan and Claudia used, COM-B, which says people need the capability, the opportunity and the motivation to change how they work.

Bonus: your next hire might be an agent

Josh Bersin gave the most provocative talk of the week. His vision of HR in 2030 is a network of AI “superagents”, each looking after one area of HR, such as hiring, learning or performance, and working with the others.

His main point was to manage agents like people. They need goals, guardrails, limited privileges and monitoring, and the narrower their job, the better they perform. Treat an agent like a new hire, in other words: give it a clear job description and check how it’s doing.

He was also optimistic. He sees AI as a technology that gives people more to work with, and he believes we’re at the bottom of the J-curve, just before the value starts to show.

We’d add one thing. If agents need clear roles and oversight, the people designing and managing them need exactly the skills discussed above: judgement, critical thinking and ethics. The more digital workers an organisation has, the more its people’s grey-zone skills count.

The most human moment

The last word goes to Louis Theroux, who has spent 30 years documenting the extremes of human behaviour. His advice for HR was simple: be curious. “Curiosity is the best antidote to shock and outrage.” Asked whether AI could ever do his job, he said not with today’s technology, because what’s missing is human connection.

In a week dominated by AI, it was a good reminder of what it can’t do.

How we can help

Many of these challenges come down to judgement and behaviour change more than technology. That’s where psychometrics and business psychology make the difference. We can help you:

  • Keep role requirements current and future-focused with expert-led job analysis and competency framework design.
  • Measure the skills in the grey, such as judgement, adaptability and critical thinking, using validated psychometric assessments, including our ready-made AI-readiness blended assessment.
  • Make your decisions fair and defensible with independent fairness analyses.

If any of this sounds familiar, get in touch. We’d be happy to talk through what it could mean for your organisation.

Get

Isn’t it time that your company gets the tools to hire the best?

Get in touch with our sales to learn all about our solutions.

Follow

Would you like to have our content delivered to your feed? Follow us in your favorite channel!

Or subscribe to our newsletter
Curious about our assessments?

Download sample reports and see how data-backed insights can help you hire better.

Want to check out a sample report to see what Clevry can uncover?