Surely We Better Than AI 1-in-20 Hit-Rate?

"Despite $30-40bn in enterprise investment into Gen[erative]AI, [our] report uncovers a surprising result in that 95pc of organisations are getting zero return"
— MIT Researchers, August 2025

This finding - albeit seemingly from a questionably small and opaque sample size - coincidentally landed just as ChatGPT5's release was widely lambasted as a 'downgrade' through slow, inaccurate, unusable responses.

Combining for many observers to talk of bubble bursting horizons suddenly rushing into view.

Taking the low mark of that range, five percent of thirty billion is a 'mere' one-and-a-half billion. It would have been nice to know exactly what glory emerged from that spend. Not just the returns - after all, wouldn't those paraded as unusually successful normally be super-keen to wallow in the limelight? - but the specific nature of said projects. Lest we replicate those hitherto doomed.

Regardless of whispered flaws behind the headline, dutifully recalling IBM's recent figures suggesting only 16pc of projects earn wider rollout, then this feels a comparison we ought be bouncing off.

Surely we have hard numbers to trumpet?

Projects of ours where we have generated solid, demonstrable, even enviable returns.

Sweeter still should we be in new product mode.

'MIT find AI has a 19-in-20 fail rate; yet we succeed every time...'

Indeed, it could well be that what we supply benefits from being part of a clearly set-out plan with the tacit backing of all involved aimed at achieving something concrete. Which if such survey stats are to be believed, seems to have eluded the fad-driven race to jump aboard the AI bandwagon.

Know your numbers, and make them known within your prospects. Even those claiming a certifiable AI project leap.