A new round of research into corporate AI use has turned up an uncomfortable number: most companies have adopted AI tools, but far fewer can point to a measurable financial payoff. Adoption rates are high. Return on that adoption is not.
This isn't the first study to find this gap, and that's the part worth paying attention to. Over the past year, multiple research groups — including academic teams at MIT and various consulting firms — have published versions of the same finding: companies are spending on AI tools faster than they're redesigning the work those tools are supposed to improve. Subscriptions get purchased. Workflows stay the same. The tool gets bolted onto an old process instead of replacing it.
The pattern consistently traces back to implementation, not capability. Companies that saw gains tended to redesign a specific process end-to-end around the tool — rethinking who does what, in what order, with what handoffs — rather than handing employees a chatbot and hoping productivity improved on its own. Companies that saw little or no return tended to treat AI as an add-on: a new app in the toolbox, not a changed way of working.
There's also a measurement problem baked into this. Many businesses never defined what success would look like before rolling a tool out, so there's nothing concrete to check results against months later. Time saved on drafting emails doesn't automatically show up as higher revenue or lower costs unless that saved time gets redirected toward something that generates value — a redeployment that has to be managed on purpose. It mostly doesn't happen by default.
For small businesses, the dynamic is especially sharp because stakes and resources are both smaller. A large company can absorb a year of tool spending with no clear return and barely notice. A ten-person shop paying for four or five AI subscriptions across writing, scheduling, customer service, and bookkeeping is spending a meaningful chunk of its tech budget on tools that may or may not be changing outcomes.