The Practice Gap
The Federal Reserve Bank of Atlanta published a number in May that should concern anyone running a business. Top AI-investing firms spend $2,800 per employee. The median spends $200. Fourteen to one.
The reflex is to write a bigger check. That reflex is wrong.
IBM found 85% of employees now have access to AI tools. Only 25% use them regularly. Writer’s enterprise report says 84% of companies have not redesigned a single job or workflow around AI. They bought the software. Nobody changed the work.
I wrote about this today because the pattern keeps surfacing in every engagement I touch. The companies generating real returns did three things before they scaled their budget. Someone in each department owns the AI workflow — not a committee, a person. At least one production process was rebuilt, not just piloted. And they measured output change, not login counts.
None of that costs $2,800 per employee. All of it requires a decision most organizations keep deferring. Meanwhile the firms that already made it are compounding — every new model release, every capability improvement lands on muscle they built last quarter. The unpracticed side stays flat.