Almost half of the organisations we surveyed (47.6%) say their current stage of AI adoption is “experimental” or “pilot”, and our experts say this is more a sign of progress than it is of failure.

Despite the hype, AI isn’t a plug-and-play solution. The jump from running a prompt in ChatGPT to deploying AI in a secure, production-ready environment is huge. As Jack Jorgensen, Data, AI & Innovation General Manager at Avec, puts it:

“There’s a big difference between punching in a search query and building something deterministic and robust enough to run in production systems.”

In other words: it’s easy to experiment with AI, but much harder to operationalise it.

The pilot phase: What’s really going on

When AI went mainstream, business leaders rushed to explore how it might improve productivity, automate tasks, and reshape work. But most quickly hit a wall. Why? Because the magic wears off when you move from ideation to implementation.

Jack assures business leaders, “Having organisations stuck in that pilot stage isn’t a bad thing. It means they’re going out and finding the limitations of the technology and where it can be applied really well.”

This experimental period isn’t just about proving AI works. It’s about learning:

• Where it doesn’t work
• Where your data isn’t good enough
• Where processes aren’t ready, and
• Where your people need upskilling

This discovery stage is critical to uncover what needs fixing before scaling, and will help businesses avoid wasting time and budget building the wrong thing.

The risks of skipping this step

In the rush to “not fall behind,” some organisations are pushing AI into production too fast. That often leads to:

• Tool sprawl and shadow AI
• Security breaches (like the now-infamous CRM upload into ChatGPT)
• Oversold outcomes with underwhelming results, and
• Burnt-out teams working with systems they don’t trust or understand

As Jack puts it, “If you’re jumping in without looking, you’re probably going to break your ankles on the way into the pool.”

The smart move is to slow down, run your pilots, and get clear on the problem you’re trying to solve.

What good looks like in the experimental phase

Here’s what leading organisations are doing right now:

• Running small pilots with clearly scoped outcomes
• Auditing internal AI usage to assess risk and opportunity
• Building foundational data and security infrastructure
• Educating teams on prompt design, ethics, and governance
• Documenting learnings to shape future strategy

This time is your opportunity to build AI capability without breaking things.

Final thought: Pilot is not a plateau

Staying in pilot mode doesn’t mean you’re behind. It means you’re taking it seriously. Rushing to production with shaky data, no security posture, and no clear goals? That’s what real failure looks like.

As JP Browne, Practice Manager from our parent company Talent, states: “Pilot mode isn’t the problem. It’s the companies skipping this step who are going to run into trouble.”

If your organisation is experimenting with AI right now, you’re exactly where you should be.

Want to find out what else our AI survey revealed? Access the full report.

If your business is ready to kick off a data, AI or innovation project, drop a message to Jack’s team.