AI Is Making Change Cheaper. How Should Governance Respond?

AI is changing the economics of digital transformation by making it cheaper to build, test ideas and learn what works. That raises a question for organisational governance: when an experiment is small, safe and reversible, how much certainty should we demand before allowing it to begin? The opportunity goes beyond faster delivery. It is about adapting decision-making, funding and assurance so organisations can learn more quickly while keeping scrutiny proportionate to the consequences of getting something wrong.

One of the most interesting things AI is doing is making change cheaper.

We talk a lot about productivity: doing the same work faster, producing software more quickly, automating tasks that previously took people hours.

But I wonder if the bigger organisational consequence is that the cost of trying something, discovering it doesn’t work, and changing direction is falling dramatically.

A lot of the governance in large organisations was designed for a world where change was expensive.

If building something required a significant investment of money and people, it made sense to do a lot of work before committing. Develop the business case. Specify the requirements. Agree the benefits. Secure the funding. Approve the design. Then build.

There’s nothing inherently wrong with any of those things. They’re rational responses to expensive decisions.

But what happens when testing the proposition costs a fraction of what it used to?

Increasingly, we can build enough to learn before we know enough to justify building the whole thing. We can put something in front of users, discover which assumptions were wrong, change it and try again.

That changes the economics of certainty.

Perhaps the question before an experiment shouldn’t always be “Can we prove this will work?”

Sometimes it should be “Can we find out whether this works cheaply and safely?”

That doesn’t mean removing governance. Some decisions remain expensive, consequential or difficult to reverse. Anything affecting people’s rights, safety, money or access to important services deserves appropriate scrutiny however cheaply we can build it.

But a small, contained and reversible experiment probably shouldn’t have to navigate the same decision-making machinery as a multi-year investment.

The governance question therefore becomes more interesting: how do we make the cost of deciding proportionate to the cost and consequence of being wrong?

That may mean clearer boundaries for experimentation, smaller funding increments, faster decisions, explicit limits on blast radius, and much clearer points at which an experiment becomes a service and requires a different level of assurance.

AI is making it cheaper to build. But perhaps the bigger opportunity is making it cheaper to learn.

If that’s true, organisations that keep governing every change as though changing direction is enormously expensive may miss a significant part of the opportunity.