Why your AI roadmap should have three items, not thirty

A thirty-item AI roadmap feels thorough and looks impressive in a boardroom, and it's almost never executed past item four. The length is usually a symptom of not having made the hard prioritization decisions: everything got included because nobody wanted to be the one to cut it.

A three-item roadmap forces a real conversation: which of these actually moves a number the business cares about this quarter, which requires infrastructure that doesn't exist yet, and which is genuinely a "someday" idea being disguised as a near-term priority to make the list feel bigger.

The prioritization framework I use is deliberately blunt: does this touch revenue or cost directly, is the data it needs already accessible, and can it ship in under eight weeks. Anything failing two of three moves to a "later" list that's honest about being later, instead of a roadmap pretending everything's imminent.

A short roadmap that actually gets built beats a long one that becomes a slide nobody opens again. The discipline isn't in the ambition: it's in the willingness to say most of the ideas aren't this quarter's ideas.