Businesses have adopted AI at astonishing speed. In the UK alone, the share of organizations using it jumped from around 12 percent in late 2023 to roughly 35 percent this year. And yet the productivity revolution everyone was promised keeps failing to show up in the numbers. A cluster of new research this week suggests the reason is less mysterious than economists have feared. The gains are being lost in two places the models never touch: the human will to finish the work, and the organizational plumbing between a good draft and a shipped result.
The most striking finding comes from Brookings, which reports what it calls the first causal evidence that attributing creative work to AI changes how people behave. In a preregistered experiment run on nationally representative samples in the United States and the Netherlands, participants were shown the same set of campaign slogans; half were told a marketing professional wrote them, half were told AI did. Those in the AI group rated the identical work as less creative, found the task less meaningful, and were 3.4 percentage points less likely to bother contributing an idea of their own, a 13 percent drop against the baseline. When people believe the machine did the interesting part, they quietly disengage from the rest.
That disengagement collides with a second problem: the work does not end at the draft. A survey of more than 300 enterprise marketing leaders by Knak found that 85 percent had missed at least one campaign launch date in the past year, and the causes were overwhelmingly operational rather than creative, led by approvals and cross-team coordination. AI has reached the first-draft stage for most teams, but 88 percent say its output still needs moderate to substantial human editing. Producing a single email routinely involves four or more people, three to five tools, and two or three rounds of revision. AI got in the door; it did not get to the finish line.
Step back far enough and the pattern holds at the whole-enterprise level. A KPMG study of 1,750 transformation leaders across 20 countries found that despite record spending on AI, cloud and automation, only 14 percent consider themselves top performers. Its conclusion is that competitive advantage no longer comes from technology at all, but from an organization's ability to orchestrate people, processes, data and governance as one. Bolt AI on top of fragmented systems and disconnected teams, and you get faster reports and prettier emails, not a transformed business.
The common thread across all of it is uncomfortable for anyone selling AI as a shortcut. The technology is genuinely fast at the parts that were never really the bottleneck, the ideas and the rough drafts, and close to useless at the parts that are: sustaining human motivation, clearing approvals, and knitting fragmented processes together. There is a real risk hiding underneath, too, as workers lean on tools they were never trained to question and drift toward accepting whatever the model hands back. The fix, then, is not a better model. It is organizational: design work so people stay invested in the outcome, and rebuild the production line between draft and launch. Otherwise teams will keep buying tools that get them to a draft and leave them stuck just short of done.