A useful stress test for any significant AI investment proposal is asking whether the business case still holds if current AI capability simply stopped improving from this point forward - no further model improvements, no further cost reductions, just today's technology indefinitely. Investments that only make sense assuming continued rapid improvement are making a meaningfully riskier bet than their proposals typically acknowledge.
Why this test matters
A substantial share of AI investment cases implicitly assume the technology will keep improving fast enough to eventually justify a return that current capability doesn't yet support - a reasonable historical bet given the pace of progress so far, but a bet nonetheless, and one that isn't always made explicit in how the investment case gets presented internally.
What passing this test looks like
An investment that delivers a clear, positive return using only today's model capability and today's costs, with any future improvement representing pure upside rather than a required assumption, is a fundamentally more defensible bet than one that only clears its return threshold if specific future improvements materialize on a specific timeline.
Why this distinction gets blurred in practice
Vendor sales conversations and internal advocacy for a given AI initiative both have real incentive to frame future improvement as a near-certainty baked into the projected return, rather than clearly separating what today's technology can deliver from what future improvement might additionally unlock.
What this means for evaluating any specific proposal
Asking explicitly which portion of a projected return depends on today's capability versus assumed future improvement, and treating the latter as genuine upside rather than a load-bearing assumption, produces a meaningfully more honest picture of the investment's actual risk.