As a co-founder of an AI startup — Algoverde.ai — I have now worked on AI adoption with leaders in large organizations in Automotive, CPG, and Insurance. From my own experience, a few truths stand out. The biggest stumbling block? Treating AI as a side project.

To me, at this point it’s pretty clear that the real challenge with AI adoption is not technology, it’s change management 101.

When companies fail to assign explicit responsibilities or tack on AI to someone’s existing workload, momentum dies.

Success requires clear job assignments, scheduled time for learning, and leaders willing to prioritize adoption. It also helps to make access to new AI tools feel earned, not just handed out to everyone regardless of buy-in.

My own experience lines up with the findings from recent Harvard Business Review research. The pattern is striking: even with widespread access, most employees only scratch the surface of these powerful tools. The problem lies in how organizations integrate and measure its use. Tracking login frequency or prompt counts won’t tell you if AI is fueling better judgment or real productivity.

So, what actually moves the needle? Here’s what works, in my opinion and in plain language:

Do’s:

• Assign AI as a real part of someone’s job description, and block the time for it. When people have the hours and the mandate they engage meaningfully.

• Make AI tools feel valuable. Limit access initially; people treat privileges with respect.

• Train your leaders first. If managers don’t understand how to use AI day-to-day, the rollout won’t stick.

• Celebrate wins across departments: sharing success stories keeps the energy high and spreads the learnings wide

Don’ts:

• Don’t let meetings get canceled or rescheduled: pilots drift without clear support

• Don’t spread your efforts too thin with a bunch of disconnected pilots. Pick a few key workflows and rethink them from the ground up.

• Avoid handing out licenses to folks with no time or little interest — they won’t move adoption forward.

• Don’t measure progress by usage stats alone. Focus on how people are applying AI to real, complex work.

• Don’t treat AI like a vending machine. It’s not an “in-out” paradigm. The best users iterate and treat AI as a thinking partner.

Both my own experience and the latest Harvard Business Review research show that organizational habits matter. You cannot treat artificial intelligence adoption as a secondary initiative.

When leaders ask employees to figure out the new tools on top of their existing workload, adoption stalls. But when they formalize the expectation and protect the time required to meet it, they signal that it’s a core strategic priority.

I’ve also noticed that managers cannot credibly advocate for tools they do not really understand. That’s why practical training is so important, at all levels. Implementing a strict “use it or lose it” mandate can help send a signal that AI access is precious.

Leaders often obsess about metrics like frequency of use, which I believe is a deeply misleading proxy for productivity. You get the most out of AI by treating it as a reasoning partner, not a quick Q&A system.

And there is nothing like explicit recognition company-wide to validate the behavior you want to scale. So give your AI champions the credit they deserve.

In the battle for attention, what other best practices have you found to ensure AI stays on top right now?

Elisabetta Ghisini is CMO and Co-Founder of AlgoVerde, an AI-powered innovation platform.