Most leadership teams are trying to be responsible about AI. They want clearer rules and a clearer ROI story before they commit, guest article by Fahed Bizzari .

The problem is that this assumes something that isn't true. It assumes AI adoption can be paused until certainty arrives.

It can't.

McKinsey's 2025 State of AI survey shows how far this has already moved. Regular AI use is now widespread. But most organisations still haven't built AI into their day-to-day workflows in a way that changes results across the whole business.

So the real question for 2026 isn't "Are we adopting AI?" It's "Can we stand behind how AI is being used across the organisation?"

Usage is not capability, AI Empowerment

Most organisations now have AI activity. People draft faster. They summarise documents. They brainstorm. They speed up routine communication.

That's useful. But it isn't capability.

Capability is when AI-assisted work stays reliable across teams and still holds when things get busy. People aren't guessing. They know what they can use AI for. They know what needs checking. They know who owns the final answer.

When capability is missing, the pattern is predictable. One team gets strong results. Another team produces sloppy work. Some people disclose AI use. Others keep it quiet. Learning stays scattered instead of spreading.

This is one reason "AI use is high" can sit alongside "impact is limited." The tools are present, but the way of working hasn't stabilised.

Why "waiting for clarity" backfires

The most common objection sounds sensible: "We can't prioritise this until ROI is clearer and regulation settles."

But if people are already using AI-and they are-then waiting mainly delays one thing: use you can rely on.

What tends to happen is simple. People still use AI to meet deadlines. They just share it less when rules feel unclear or political. They stop asking for permission if the approved route is slow. They learn privately, not together. Over time, leaders lose visibility just as usage spreads.

IBM's CEO study release points to the same issue. As CEOs push to scale generative AI, they report that the hard problems are workforce, culture, and governance—not just the technology.

So waiting doesn't reduce risk. It often makes risk harder to see.

Responsible AI only works when people can do it

At this point, many organisations respond with policies and guidance. Some of that is necessary. But documents don't run the business. People do.

If you want responsible AI use to hold in real work, it has to be something teams can actually execute. That means clear boundaries, realistic checking, clear escalation paths, and clear ownership.

NIST's AI Risk Management Framework is useful here because it treats risk management as something organisations implement and maintain, not just something they write down. NIST also published a Generative AI Profile as a companion resource, reinforcing the same practical message: generative AI creates distinct risks, so organisations need concrete practices that work in context.

So the leadership test isn't "Do we have a policy?" It's "Do we have a way of working that still holds on a busy Tuesday afternoon?"

The 2026 priority is an approved pathway

This is what I mean by AI Empowerment.

AI Empowerment is how you build an approved pathway. It is the human and organisational ability to use AI in work that is safe, repeatable, and accountable. Then you make that the default way people get work done.

An approved pathway removes guesswork. It answers a few basic questions:

Where is AI encouraged, and where is it not?

What must be checked, and how much checking is enough?

Who owns the output when AI contributes?

What happens when someone is unsure, or when the stakes are high?

When those answers exist and are reinforced, two things happen at once. People move faster, and leaders gain visibility. Good practice becomes easier than workarounds.

This is also the responsible answer to the ROI-and-regulation objecti...