This is the third installment in a series on what it takes to close the gap between AI investment and tangible business impact.

In the first two pieces, I argued that AI adoption is fundamentally a people challenge – not a technology one. And I walked through why normalization – that initial work of building psychological safety and making AI feel less like a mandate from above – is where adoption begins.

But normalization is foundational, not sufficient for scaling and sustaining a transformation.

Listening builds trust, transparency creates permission and leaders modeling curiosity make space for people to be curious too.

Yet at a certain point, readiness must turn into action, and trust must move into experimentation – and that’s where things often stall for organizations.

What experimentation entails

Experimentation is not a single workshop or training session where people learn AI in theory and hope to apply it later.

Experimentation is the deliberate work of creating low-stakes opportunities for people to try AI with daily tasks, see what happens, share what they learn and let those discoveries shape how their team and the broader organization evolves.

Why experimentation stalls

Many organizations are making the mistake of treating experimentation like compliance.

They may schedule training, mandate it or expect it to happen in a structured, predictable way. Then they’re surprised when adoption remains stuck at the edges – concentrated among early adopters while the rest of the organization watches, wondering why this whole “AI thing” isn’t for them.

What tends to move people from curiosity to confidence is relevance. When people see how AI fits into their work, adoption stops feeling like a mandate being forced upon them and starts feeling like a tool they’re better off with than without.

Who influences experimentation

Experimentation doesn’t happen without active support – and managers’ involvement is often what determines whether it takes root or trails off.

Microsoft’s 2026 Work Trend Index found that when managers visibly use AI themselves – not just endorse it – employees report a 17-point lift in AI value, a 22-point lift in critical thinking and a 30-point lift in trust in AI tools.  

Employees who strongly agree their manager actively supports their team’s use of AI are 8.7x more likely to say AI has transformed how work gets done. 

Manager visibility and support are the prerequisites for experimentation to become embedded in how work gets done.

And beyond that non-negotiable support, companies can deploy three strategies to accelerate experimentation:

1. Design low-stakes opportunities to try, not mandatory programs.

The difference between training and experimentation is permission to fail.

Experiments are designed to surface discovery. The expectation is learning – which includes failure. The point is to uncover insights that shape what comes next, not to prove mastery.

This changes how people engage. Instead of one-size-fits-all training, create role-specific challenges.

These don’t need to follow a single format.

Some organizations run week-long sprints where teams tackle a specific workflow problem with AI.  

Others build 15-minute “AI challenges” into team meetings – quick, low-pressure moments where teams tackle something together and debrief in real time.  

At FleishmanHillard, we’ve deployed “try this” email campaigns that highlight role-specific tips and best practices – paired with reinforcement in team meetings – and structured, cross-functional hackathons and competitions where groups solve real workflow problems with AI.  

Format matters, but less than the regularity with which you encourage and provide opportunities for your people to try something with their work, see what happens and reflect on it – moments where stakes stay low, the learning gets documented and momentum builds as people see peers discovering things that work.

2. Spread learning through peer voices and stories, not polished case studies.

Your AI wins will get turned into case studies – charts, metrics and messaging locked in to prove ROI. While these matter for leadership dashboards, they often read less like something a peer figured out and more like something the company or an expert achieved.

Peer stories work differently. They come from someone familiar and in a similar position. They’re messier, they show what someone was really thinking when they tried something and they make room for context – “Here’s where I am, here’s what I tried, here’s what happened and here’s what I’d do differently.”

That messiness is what makes them powerful. It signals that perfection isn’t the bar – and if someone who thinks like you and works like you figured something out – suddenly that same experiment feels possible for you too.

Those stories create permission in ways polished case studies never do – which is why leaders should find ways to share these.

This could take several forms.

Create internal campaigns where teams share what they tried that week and what they learned – misfires included – via Slack threads or Teams channels.  

Host show-and-tell sessions where someone walks through how they solved a problem, where they got stuck, what went wrong – and invites the room to help troubleshoot next steps.  

Or establish dedicated architect and ambassador roles – like we’ve done at FleishmanHillard – where builders and super users experiment alongside their teams, share what’s working and what isn’t, and create permission for others to do the same. 

And with peer stories, tone is everything. “I tried this and it didn’t work, but here’s what I learned” is infinitely more relatable than, “Here’s how our company is transforming productivity and driving efficiency.” One invites personal experimentation – and the other signals compliance.

3. Build informal peer-to-peer momentum instead of formal training.

Training is periodic and linear. Peer-to-peer learning is fluid, constant and evolves as your organization does.

Both are valuable, and you should deploy each as needed, but peer-to-peer learning builds adaptive capacity – the kind that compounds and grows alongside your culture.

When you create simple mechanisms for ongoing peer-to-peer learning – “What I learned this week” rituals in team meetings, Teams threads where people drop quick tips, debrief huddles where someone walks through how they applied AI to a real use case, side conversations where a peer shares a shortcut – learning stops being something that happens to people and starts being something they do together.

At FleishmanHillard, we’ve made this easier by building off-the-shelf training resources and templates that any role can use or adapt for their teams – simple scaffolding that removes friction and makes peer sharing more accessible.

We’ve witnessed firsthand that those moments compound, and they reshape how your organization thinks about discovery and experimentation.

They also serve as a continuous feedback loop. You may learn more about what people care about, what confuses them and what would help them in a few weeks of informal conversations than from your annual survey.

On top of all that, assuming you’re following through on what you hear, your people will feel like their voices shaped what comes next – because they did.

How you know experimentation is becoming part of your culture

Experimentation is a continuous, messy, non-linear process – not a single moment. Here are three signals you’re heading in the right direction:

AI is being applied to everyday work. Analytics and team check-ins show employees testing AI with real tasks.  

Learnings are being shared. Examples, wins and failures surface in meetings and peer showcases. The conversation has shifted from, “how do I use this?” to, “here’s what I tried and here’s what I learned.”  

Confidence is building. The tone in surveys shifts from, “I’m not sure where to start” to, “I’m figuring out where this could help.” People are getting more and more curious and taking small, concentrated risks because they feel safe doing so. 

Building the cultural conditions for experimentation

Right now, organizational culture is roughly twice as powerful as individual mindset in determining whether AI delivers value.

The organizations accelerating adoption are the ones making room for people to learn and figure out what this technology means for their work – where people feel safe trying them, failure is a learning opportunity, peer discoveries shape strategy and use becomes personal enough to stick.

So, the question for leaders right now is less about the technology itself and more about whether you’ve created the conditions for everyone to use it. And if you haven’t normalized this shift and built experimentation into how your organization operates, the answer will always be no – no matter how good the tools are.