Ask people who now lead AI work how they started and almost nobody describes a strategy. They describe an annoyance.
The pattern, repeated
Step one - a small, irritating problem. Not a transformation programme. A report that took three days. A backlog of enquiries nobody could triage. Something they personally found tedious.
Step two - an off-the-shelf tool and no permission. They tried something on their own work, in their own time, with no budget and no announcement.
Step three - a measurement. This is the step that separates the leaders from the enthusiasts. They knew the before number and could state the after number.
Step four - they told someone who mattered. One slide. Problem, approach, result, what it would take to do properly.
Step five - they were handed the next one. Scope arrives when someone has demonstrated they can be trusted with it.
What they did not do
They did not learn to code first. They did not wait for a strategy. They did not ask for a budget before having evidence. And they did not start with the most important process in the business, which is the fastest route to a visible failure.
Where training fits
Most describe formal training as arriving in the middle, not at the start - taken once they had a real problem and needed structure, vocabulary and a governance frame to scale beyond a personal experiment. That is the point at which Leading AI Innovation & Organisational Transformation does its work.
The honest caveat
This is survivorship bias territory. Plenty of people ran the same play and hit a wall on data quality, procurement or a manager who was not interested. The pattern raises your odds; it does not guarantee anything.
The bottom line
Small problem, cheap tool, measured result, told the right person. Then train properly to scale it. Related: starting small with AI and how to build a business case for AI.