HR leaders will often focus on engagement as a priority, but in the age of AI when change is a constant, should focus shift to experimentation instead?

For Amy Edmondson, Novartis Professor of Leadership and Management at Harvard Business School, the answer is clear.

“The cycle times are shrinking. There’s always been change, there’s always been learning, but the speed of innovation has obviously never been higher. The human brain can’t easily keep up with that,” she says.

At UNLEASH America 2026, Edmondson delivered a keynote address on how HR leaders can study and embrace failure – or what she terms “intelligent failure.”

Sitting down with UNLEASH for an exclusive interview, Edmondson laid out a challenge for HR: stop over-indexing how people feel about work, and start designing systems that accelerate learning.

“We have to help people get more comfortable with experimenting, and our natural inclination, especially in work and in a hierarchy, is to only do things we know in advance will work,” she explains.

“That is a recipe for getting left behind, both individually and organizationally.”

Organization control needs to make way for scaffolding to create the right environment

HR leaders currently face an explicit decision: continue to optimize for stability and predictability, or accept that learning, including failures, needs to happen continuously and at scale.

One of Edmondson’s most important messages for HR is that traditional, top‑down control systems are no longer fit for purpose when aspiring for employee-driven learning and continuous adaptation.

“What’s needed to operationalize some of these conceptual ideas, is systems and practices that act as scaffolding that help people do it,” she states.

While Edmondson says she is “agnostic about how they do that,” there is a definitive need for “having conversations about what we should try next, then thoughtfully assessing it so we can pick what’s the next best thing to do.”

Ultimately, this puts HR squarely at the heart of organizational design: building the guardrails and guidelines that allow and encourage teams to ask the right questions, surface what they know, test ideas, and learn from outcomes.

However, this also means organizations will be “even more vulnerable to the performance of teams,” Edmondson adds.

“You can no longer really expect to have an organization that’s top-down where everything just cascades nicely through the organization,” she explains.

It’s about being much more of a complex adaptive system, where you have teams over here doing something, teams over there doing something…all of which adds up to positive developments.”

AI means workers are choosing ease over effort, and HR needs to get away from productivity mindsets

Amy Edmondson, Novartis Professor of Leadership and Management at Harvard Business School

The increasing use of AI is raising the stakes, as evidenced by the rise of ‘work slop’.

Edmondson says this is a “beautiful example of a lack of systems thinking” but heightens the misdirected focus on productivity that AI has exacerbated.

“Why are we talking about productivity? It’s an old-fashioned metric. It refers to a measure output over effort – which only makes sense for tasks and processes that are clear in advance,” she states.

“We should be talking about quality, or value, or which customers or markets we’ve helped solve which problems.”

A fundamental issue AI has raised when it comes to learning, and specifically learning from ‘intelligent’ failure, is that it offers workers an easier option for task completion than before.

“One thing we know is that people choose ease over effort; yet want them to choose effort over ease,” Edmondson tells UNLEASH.

“I do not mean, of course, we want them to find a harder way to do the same task. I mean that we want people to be willing to do hard things – not just easy things – and to experience the satisfaction and even joy seeing the results they can achieve.”

The task in front of HR leaders is getting employees to take the harder path, and Edmondson advises that to do so they firstly need to understand what’s being asked and what the need is – essentially, the purpose of the task.

While there’s no single answer to this, developing organizational design that supports AI use as genuine experimentation will be a differentiator.

Edmondson recommends engaging workers using AI “like scientific leaders do in the hard work of genuine knowledge creation, which is not the same as just giving them the tool and saying, ‘go play’.”

Psychological safety and leading like scientists will be keystones of success for HR leaders

Paramount to creating the right environment for experimentation is psychological safety – when workers feel safe to speak up, admit mistakes and challenge authority without the fear of reprisal.

While many organizations claim to support innovation and intelligent failure, Edmondson is more realistic: “Psychological safety is not prevalent; it is not the default.”

Not having this doesn’t mean organizations are broken or toxic, she adds, it’s just the default.

“We humans have very predictable, normal reactions to hierarchy; we engage skilfully in impression management,” Edmondson explains.

It takes work to build a culture that appreciates smart failure and is psychologically safe for candor. They don’t happen automatically; they have to be built and cultivated.”

HR leaders are once again at the center of this – building the right leadership behaviors that facilitate psychological safety at scale.

Core to this is acknowledgement that leadership doesn’t have all the answers, particularly during periods of consistent change and disruption.

For Edmondson, this means leaders being truthful and showing “humility about the situation that you find yourself in, which is that it’s a brave new world.”

“It’s hard to learn if you already know, and I’m sure too many leaders have this training or belief that you have to say: ‘I know. I have all the answers. I’ve never been wrong in my life’.”

She also reiterates the importance of leaders and managers adopting a similar approach to scientific leadership to succeed at experimentation: focus on asking the right questions rather than pretending you have the answers.

“They’re keen to discover, not just to produce, because they understand that our future value is going to be the result of our discovery rather than the result of simple execution.”

For HR leaders, Edmondson has one key piece of advice for the AI-enabled future of work: committing to “structured experimentation” rather than “letting 1,000 flowers bloom.”

Organizational advantage won’t come from doing more work, but from learning faster than everyone else. HR now has to decide whether it is ready to lead that change.