AI’s impact on hiring, decisions and culture: What’s left unspoken

The interview process changed, decisions were moved into a prompt, and the agreement with employees was rewritten. Here is what nobody is saying and what to do about it.

Let me start with a picture.

A woman walks into an interview. She is a senior professional, so she has done this before. She prepared as we were all taught to. She has her case studies ready, examples of problem-solving and impact, the product launch, and anecdotal data about the team she rebuilt after a layoff. But during the interview, no one asks her for any of this.

What she gets instead is a laptop, a real problem the organisation is working on, an AI assistant and 40 minutes. The draft the tool gives her contains a mistake. The interviewers across the table are not listening for her problem-solving, impact stories or even her wins. They are watching how she responds when she finds the mistake, and whether she finds it at all.

That is called an observable interview, and I believe it is critical for us to talk about because a version of that scene is already happening inside some of the largest organisations in the country, changing culture in ways that few people are saying out loud.

What was said on the AIQ summit stage

Fortune held its first AIQ Summit this fall, and the headline in its coverage read “AI has killed org charts.” I read the whole piece, and what the chief people officers actually said was more interesting than the headline.

Stacey Dillow at HPE said the technology was never the hard part. The hard part is rethinking how the work gets done. If you drop AI into a process you have not changed, you get the same answer you always did. Danielle Gonzalez at Palo Alto Networks said they now hire for curiosity, agency and the ability to unlearn, and they test for it through observable interviews and hackathons, with and without AI. She also said employees who used to call her or their HR business partner now ask an internal agent instead. And Drew Holler at Lennar said the organisation will give people the tools and the training, but upskilling is the employee’s responsibility. (Preston Fore, Fortune, Oct. 1, 2026.)

None of them said the org chart is dead. What they described is a new way of working between organisations and the people who work for them. The old “agreement” within the workplace that most of us experienced went something like this: 1) We hire and train you. 2) You stay with the organisation. 3) Over time, you earn experience, skills, seniority and standing. The new agreement looks like this: 1) We hire you and hand you tools. 2) You upskill yourself and prove you can adapt. 3) You keep proving it, starting before you are even hired.

What was not said

No one on the AIQ Summit stage said that Holler’s sentence is a contract. But it signals a noticeable shift in the expectations we hold for talent and in the employee life cycle.

What happens to decisions when AI replaces the team responsible for people and culture? When an employee asks an agent instead of calling their HR business partner, the question is answered faster (yes), but the conversation then disappears. That conversation was where a manager noticed someone on their team was struggling. It was where a new hire learned who to trust during onboarding. Decisions that used to happen between two people now happen inside a prompt, and nobody is tracking what gets lost in translation. Furthermore, no one mentioned what the shift in the interview process itself does to a culture. That is the part I know best, so let me spend some time here.

What these quiet changes do to culture

I spent years in hiring rooms at organisations where culture was my job, and here is one of the unspoken rules that I learned in those roles. Whatever you test for at the door is what becomes the norm at the organisation. If you change the interview, you change the culture, whether you planned to or not.

The observable interview moves four things at once, and each one deserves to be named plainly:

  • Hiring moves from telling to showing. We used to be hired on the account we gave of our work. Now, we are hired on a live sample of it, performed in front of strangers, on a clock. Watching someone work does tell you more than hearing about it, and I believe that. But some people who did the work beautifully cannot perform it on demand in a room full of evaluators, and if you lose them, you are measuring composure more than capability.
  • Development moves from the organisation to the employee. The learning budget stayed where it was. The responsibility moved to the person with the least leverage and the least time. The organisation still pays for the course; the employee now has to find the hours in a calendar they do not control and prove the result.
  • Clarity moves from the job description to the manager. When work is constantly being redesigned, a job description cannot keep up, so the question “What is my job now?” must be answered by a person—the manager. Where no manager answers it, people fill the silence with anxiety, and anxious people do not experiment, which is the very trait the observable interview was designed to find.
  • Status shifts from what you know to how quickly you can adapt or flex. For most of our careers, the person who knew the most was seen as the leader and expert in the room. However, in an organisation that celebrates the ability to unlearn, the person who adapts fastest holds the status, and that can be the two-year hire rather than the 20-year veteran. That is a reordering of who has power, and it is a status threat. People feel that threat long before anyone even thinks to measure it. I will come back to what a status threat does to a brain.

 

If you are going to run an observable interview, run it honestly

The phase is new, but the method has a long history. Industrial-organisational psychologists have studied work-sample tests for decades. Schmidt and Hunter’s 1998 meta-analysis put them at the top of the list for predicting job performance, and the 2022 reanalysis led by Paul Sackett lowered that estimate and moved structured interviews into first place. Read together, the lesson is that watching people work is a real signal, and structure is what makes it a strong one.

So, here is what I would ask of anyone building one:

  • Treat the round with AI and the round without it as two separate tests. One shows you judgement. The other shows whether a person directs the tool, checks the tool, and overrides the tool when it is wrong.
  • If you plant a mistake, tell candidates you are watching how they check the work. People who know what is being evaluated focus on the problem rather than scanning the room. You get a cleaner, fairer read.
  • Use a rubric and two raters, and score how they got there more than how finished it looks. Otherwise, you are hiring for performance under evaluation, which is a real skill and not the one you need.
  • Keep it short and separate from live company work. Pay for it if it runs long, and give everyone the same tools and a few minutes to warm up. Someone who has never had access to these systems has had less exposure, but exposure is not the same thing as capability.
  • Remember that the candidate is interviewing you also. Your new agreement is on the table before anybody signs anything, and the people you most want to hire are experiencing it.

 

The part I came here to say

The word that stuck with me from that stage was “unlearn.” So, let me tell you what unlearning actually asks of a brain, because that is where the culture question starts.

When you let go of something you knew, your brain registers it as a loss. Your expertise is tied to your identity, so updating it costs you something, even on a good day. Now, add a threat to your status on top of that. The brain handles a threat to status with some of the same machinery it uses for a threat to the body, and once that machinery is running, your ability to take in new information and change your mind narrows significantly. When an organisation asks people to unlearn while those same people are busy protecting their position, it does not work because a brain in a threat state holds on to what it knows.

READ MORE: AI has not changed what good work is at Agoda. It has changed how people learn it

The same is true of curiosity. People become curious when they feel safe enough to try something new and risk being wrong. That means curiosity is something you can cultivate. So, when a candidate shows you curiosity in an interview, part of what they are telling you is that your room felt safe to them. The interview is telling you something about your culture, too.

And if unlearning needs safety, then the new agreement has a second half, and it is the half that went unspoken on that stage. When upskilling belongs to the employee, I believe the organisation owes four things in return:

  • Protected time to actually do it.
  • A clear map of which skills the organisation is betting on.
  • Permission to fail in public while they learn.
  • An honest answer to the question people are already asking each other: Where is the productivity gain going?

 

What organisations can do now

I do not want to leave you with a diagnosis and no treatment, so here is where I would start if I were sitting in your seat.

  • Write the agreement down. Put in plain language what the organisation now expects from people (upskilling, adaptability, continuous proof) and what it commits to in return (time, a skills map, room to fail and where the gains go). Then, put it where people will actually see it: the job posting, onboarding and manager training. People can trust an agreement they can read.
  • Give learning a budget of hours. The dollars are usually there, but the hours disappear. Pick a number per month, put it on the calendar and make protecting it part of how managers are evaluated.
  • Publish the skills map. Name the handful of skills the organisation is betting on over the next two years and what good looks like at each level. Nobody can upskill toward a target that has not been named.
  • Retrain managers for the job they now have. The manager’s work moved from assigning tasks to redesigning them, and from being the expert in the room to making it safe for someone not to be one yet. Train for that directly: how to run a redesign conversation and how to name a skill gap without triggering a threat response.
  • Decide which moments stay human. Decide in advance which conversations an agent never replaces: for example, a new hire’s first 90 days, performance conversations, and anything involving a career change.
  • Measure the conditions before the rollout. Trust, safety, clarity, belonging, and accountability decide whether an AI rollout lands or stalls. Baseline them before you buy the tools and again 90 days after. You cannot intervene on what you cannot measure.

 

So, here is my ask. Before your next AI rollout, your next hiring cycle or your next all-hands, answer three questions in writing: (1) What do we expect from our people now that we did not expect two years ago? (2) What have we committed to in return? (3) Does everyone in the building know both answers? If you can answer all three, you have an agreement. If you cannot, you have an unspoken one, and unspoken agreements are where cultures quietly come apart. The org chart is still there, but the agreement underneath it changed. People stay for the leaders who say the new agreement out loud and keep their half of it.


About the Author: Dr Cherise Bernard is the Founder and CEO of Modern Work Consultancy. This article was first published on HR Executive.

Share this articles!

Latest Topics

More from HRM Asia

Subscribe to Our Newsletter

Stay updated with the latest HR insights and events,
delivered right to your inbox.

Sponsorship Opportunity

Get in touch to find out more about sponsorship and exhibition opportunities.