The most important decision in any artificial intelligence (AI) project is not which technology to buy, but which problem it is meant to solve, according to Josh Tyrangiel, the journalist and author.
Tyrangiel made the case in a recent episode of the Harvard Business Review (HBR) IdeaCast, the podcast produced by the US-based management publication.
Problem before technology
He argued that successful AI adoption depends on organisational discipline: choosing a high-impact problem, executing a solution and then measuring the value it delivers.
Tyrangiel warned against what he called the "magic" narrative common in marketing and advertising for AI products.
That narrative implies that simply purchasing software will automatically boost productivity, cut costs and improve efficiency without any further work from the organisation buying it.
The magic myth
Treating AI as a magic fix, he said, is counterproductive and can set a company's efforts back rather than advance them.
Instead, leaders need what Tyrangiel described as "scalpel-like" precision in identifying the exact problem an AI tool is being asked to address.
He also stressed that technology alone cannot close the gap between a company's existing systems and new software.
The human bridge
That job, he said, falls to specific, skilled people within the organisation who understand both sides of the transition.
The HBR IdeaCast episode was framed as an attempt to give business leaders a clearer framework for deciding where to direct their AI spending.
It included practical advice on communicating change internally, winning staff buy-in and pinpointing where AI investment is likely to produce measurable results.
Lessons across sectors
Examples cited on the podcast were drawn from healthcare and other sectors, illustrating how the same principle, problem first, technology second, applies across industries.
Tyrangiel's central message is a rebuke to the assumption, widespread among executives, that AI adoption is primarily a procurement exercise.
For HBR's audience of managers and executives, the takeaway is blunt: without a clearly defined problem and the right people to execute against it, even the most advanced AI software will deliver little of value.
