$665 billion in. Not much to show for it

Enterprise AI spending will hit $665 billion globally this year. According to McKinsey’s 2026
Global AI Survey, 73% of those deployments won’t deliver their projected ROI.
Most people hear that and assume the tech isn’t working. It is. A study of 140 enterprise AI
implementations across financial services, retail, manufacturing, and healthcare found that only
23% of failures were technical. The other 77% came down to how the organisation was set up
— or wasn’t. 41% of underperforming projects had no clear business owner. 34% broke
because the AI system was designed around a process that didn’t match how the business
actually works. And 61%, according to MIT Sloan, were approved with no framework for
measuring whether they worked after go-live.
That last one is worth reading twice. More than half of enterprise AI projects got budget
approval with literally no way to prove they delivered anything.
The top-quartile organisations — the ones seeing 3x to 5x returns — aren’t running better
models. They’re running the same models inside businesses that were reorganised to use them.
Clear ownership. Workflows that were actually redesigned, not just annotated with “now add AI
here.” Governance built in before the first model went live, not bolted on after the audit.
One more. Retrofitting governance to an existing AI deployment costs 3x to 5x more than
building it in from the start. So the shortcut wasn’t a shortcut. It was a loan with interest.
Source: AI Governance Today — The $665B AI Spending Crisis