The gates are what stop speed becoming recklessness.

AI changes the speed at which organisations can explore an idea.

Things that once required months of analysis, development and testing can sometimes now be made tangible in days. People can put their hands on working technology rather than spend weeks discussing what might eventually be built.

That speed is valuable.

It also creates a management challenge.

The gates are what stop speed becoming recklessness.

The idea of gates is not new.

Earlier generations of systems development brought substantial discipline to requirements determination, prioritisation, testing, implementation and production readiness. The work could take months or years, but there was clarity about what needed to be established before moving forward.

Agile subsequently brought something equally valuable: iteration, visible backlogs, product ownership, rapid feedback and much closer participation by the people doing the work.

Practical AI adoption should discard neither tradition.

It should take the useful disciplines from both and operate at a much more compressed tempo.

A progression, not a prescription

One useful progression is:

Sandpit → Sandbox → Pilot → Scale → Production

A Sandpit creates room to explore possibilities.

A Sandbox allows promising ideas to be tested safely and closer to the real workflow.

A Pilot brings in real people and real work to reduce important uncertainties.

Scale deliberately increases operational complexity and consequence.

Production makes the capability part of how the business actually operates.

But this should not become another rigid methodology.

Organisations need multiple entry points.

A mature technology may enter directly into a sandbox or pilot. An established platform might begin with questions of scale. A genuinely novel opportunity may need to start in the sandpit.

The purpose of the progression is not to force every idea through identical steps.

It is to make the questions progressively harder as operational consequence increases.

Keep the backlog visible

Underneath the progression sits something equally important: the backlog.

A backlog makes opportunities, problems, use cases and emerging patterns visible without pretending that all of them should be pursued.

Attention should narrow.

Visibility should not.

Good ideas may need to wait. Some will become more valuable as technology changes. Others may combine with opportunities elsewhere in the organisation. Some will simply disappear.

A visible backlog allows an organisation to pursue fewer things with greater discipline without losing sight of what may matter next.

Adoption is operating work

Across every stage, several questions persist.

What happens to the workflow?

Who needs to participate?

Where does human judgement belong?

What governance is proportionate?

What value are we actually creating?

Participation matters because AI adoption should not feel as though something is happening to the organisation at extraordinary speed.

The people closest to the work need opportunities to see it, touch it, question it and understand how their judgement fits into the changing workflow.

Human judgement is not there simply to slow the technology down.

It helps manage quality, confidence and adoption while patterns are still being established and validated.

The faster AI becomes, the more deliberately these questions need to be addressed.

Speed is not the enemy of discipline.

Properly designed, the two reinforce each other.

And that may be one of the most important opportunities in practical AI adoption:

to create useful value progressively, without waiting years — and without abandoning the fundamentals that help organisations change well.