The future of AI and work: Insights from the Culture First Leaders Forum

Thanks to AI, the world of work is changing rapidly.

“We’re on an exponential curve,” explains Didier Elzinga, Culture Amp's co-founder and executive chair, in a room full of HR leaders at Culture First North America 2026, during a candid fireside chat with CEO Caroline Rawlinson.

While it may feel like others have access to more information than you do, Didier says that’s not the problem. "There is literally no information that the people inside Anthropic, OpenAI, Legora, [or anyone else] have that's not available to you." He sees the problem as one of “imagination asymmetry."

In other words, you don’t need more information. You need more imagination.

“When you're on an exponential curve, you only have to go a little bit further along for everything to look very, very different,” he continues. Instead of hunting for the one tool or prompt that will unlock everything, he believes the only way for HR leaders to keep pace with the technology is to stretch their sense of what's possible.

Here are key takeaways from the conversation about what the future of AI might look like.

Agents are coming for white-collar work, but that's not the whole story.

HR leaders are being asked to rationalize management layers and get more leverage from fewer people, while individual contributors take on a new kind of management: running agentic workflows.

Managing multiple AI agents, catching what's fallen over, and deciding what needs a human's attention is mentally demanding work that most individual contributors have never had to learn. Organizations will need to teach it deliberately, rather than expecting employees to develop these capabilities on their own.

To help, Didier encourages leaders to ask themselves questions like:

• How do I give tools to my people that allow them to do more?
• How do I help them have leverage?
• How do I create value with AI?
• What does it mean to use AI to harness what's best about being human?

What it actually means to be "AI native"

"AI native" is a phrase that gets thrown around a lot without a clear definition. Here’s how Didier explains it:

• AI-assisted organizations use AI to synthesize information, build plans, and execute faster. That's most of us, Didier notes: "Every one of us in the room is AI-assisted, not AI-native, today."
• AI-native organizations treat every interaction – with a prospect, a customer, an employee, or a system – as a chance to improve the system as a whole. The organization compounds and learns from itself continuously.

Didier says, “Why compounding organizations will win, and why learning systems like this win, is because they're learning every second.” Systems that learn get smart really fast, so the goal is to build organizations that can do that – not with computers on their own, but with human interaction.

What that looks like in practice

Didier pointed to an Australian online lending company that carved off roughly 10% of its business, its online lending arm, to become a self-contained learning loop before scaling AI anywhere else. The company's chief product officer told him it took eighteen months to get right, and that the biggest lesson was simple: You have to have eyes on glass.

Not technologists watching dashboards from a distance. You want the actual people who do the work to watch the system in motion, catch what breaks, and treat every failure as a chance to close a gap and improve the AI.

If your organization is redesigning a process around AI, the people currently doing that work need to be part of the redesign, not informed about it after the fact. "If you don't have that conversation with them, you can't co-design the process," Didier advises. Bring in outside consultants to redesign a process without the people who run it today, and you’ll lose the institutional knowledge that would have made the new system actually work.

Your values now have two audiences: humans and agents

This is the part of the conversation HR leaders in the room leaned into hardest, and it's arguably the most immediately actionable: When was the last time your company updated its values? Because those values now need to make sense for AI agents, too.

"The values that you're building to run your organization are no longer just for humans," Dider says. Take urgency, a value plenty of leadership teams reach for when they want to drive change fast. He cautions: Urgency is one of the worst values you can hand an agentic system, because it's effectively a license to cut corners.

That reframes values from an internal culture document into something closer to a context layer, informing how both people and AI systems are meant to show up, make decisions, and handle ambiguity. Get that layer wrong, and you're not just misaligning people. You're programming your agents to behave badly.

Culture Amp rewrote its own values, a set Didier built the company on 15 years ago. Caro describes a recent leadership offsite where the team spent a full day, away from screens, working through which of those values are distinctly human and worth protecting, and which no longer serve the organization.

She shares the six design principles guiding Culture Amp’s transition. They're worth considering for your own organization:

• Decisions have owners, and owners decide. The most expensive failure isn't a bad decision; it's no decision, or no clarity on whose call it was.
• Context is a shared, machine-readable asset. Hoarding information used to be a source of power. Now, the people who document and share context well, so both humans and agents can use it, are the ones worth hiring and promoting.
• Agents do real work under real authority. Agents aren't a side experiment. They support your workforce with real scope and real accountability.
• Humans do what only humans can do. Judgment, discernment, relationships, creativity: Name the things you won't outsource, and protect them deliberately.
• Everyone builds, including leaders. A role that's purely about managing people or routing information isn't sustainable anymore. Everyone needs to be creating something.
• Speed and safety together. Move fast, but not at the expense of the obligations you hold to your people, your customers, and their data.

Three things HR leaders can act on this quarter

• Audit your values for a second audience. Reread your current values and ask which ones could be dangerous if an AI system took them literally. Urgency, aggressiveness, and "move fast" language are worth a second look.
• Put eyes on glass before you scale anything. Before rolling out an AI workflow widely, start with one contained slice of the business. Have the actual practitioners, not just the technical team, watch it run and flag what breaks.
• Start small and then widen the guardrails. Caro advises beginning with the AI equivalent of simple math. Ask the AI basic questions, and once it consistently answers correctly, expand what you trust it to handle. Trust is built incrementally, not granted all at once.

How do you know the AI is actually right?

An audience member asked the question a lot of HR leaders are quietly sitting with: How do you stress-test AI output when you don't have the expertise to independently verify it?

Didier offers a practical technique for evaluating AI answers to complex questions: Use two models against each other, a version of what's technically called a generative adversarial network, or GAN. Have one model produce the work. Feed it to a second model for critical feedback.

Then, route that feedback back to the first model and repeat until the output clears a quality bar you set. Models tend to rate their own work generously, he noted, so pairing them against each other and looping the feedback back in produces sharper results than asking once and moving on.

The opportunity underneath the disruption

Didier closed on a genuinely optimistic note about what AI means for HR.

"For so long, HR was essentially a reactive firefighter," Didier says. HR was expected to manage a queue of problems people couldn't solve on their own. Now, he says, "Not only do we have all these amazing humans, we have these incredible agentic workflows that are actually pulling in as their core source input, the culture of the business, the values of the business, and the way the business should show up – and you are the steward of those things.

Nobody in the room, including Didier and Caro, has fully figured out what the future of AI will bring. That's the honest state of the industry right now. But the organizations that get furthest ahead won't be the ones with the most information. They'll be the ones willing to imagine further along the curve, and build toward that vision.

Want to stay ahead of the game?

Watch the webinar to learn other predictions and trends shaping the future.

Watch now