The boardroom conversation about AI has shifted. Two years ago, the question was whether to invest. A year ago, it was how fast. Today, it's a far more uncomfortable one: did any of it work?
Across our conversations with finance leaders, five questions come up again and again. Every one of them has a defensible answer — if the data is set up right. Most organizations' data is not.
First: how much are we actually spending? Not the line in the vendor contract — the all-in cost, including integration overhead, training time, and the shadow seats nobody remembers buying. Second: who's using it? Third: is it moving the metrics it was supposed to move? Fourth: what's the return — expressed as a number, not a narrative? And fifth, the hardest: what would we do differently if we were making these decisions again?
The organizations that handle this moment well are the ones treating AI spend like every other category of capital allocation: measured, attributed, and subject to honest retrospectives. The ones that don't are the ones still forwarding vendor decks as evidence.