The average knowledge worker is now juggling eight projects simultaneously, spending 37 percent of their time on tasks unrelated to their actual job, and 87 percent say they lack the time or capacity to coordinate with colleagues. These are findings from Atlassian's 2026 State of Teams report, which surveyed thousands of workers across multiple countries. They are not the numbers of a workforce that has successfully absorbed a productivity-enhancing technology. They are the numbers of a workforce that is busier, more fragmented, and more stretched than before the AI tools arrived.
The paradox Atlassian identifies is precise: AI has delivered on its promise of individual speed. Workers are moving faster. And yet, at the team and organisation level, work has become more chaotic. The coordination that used to happen naturally, through shared processes, common tools, and regular team rhythms, has broken down as individuals and departments have adopted AI independently, at different speeds, for different purposes, producing outputs in incompatible formats. The result is what the report calls the "fragmentation tax": for Fortune 500 companies, the estimated annual cost of uncoordinated AI adoption is $161 billion. Speed without alignment turns out to cost more than slowness with it.
This fragmentation problem has a counterpart in a separate finding: most organisations cannot demonstrate that their AI investments are actually working. Newsweek's AI Impact newsletter, drawing on interviews with business leaders across sectors, found that companies are overwhelmingly measuring activity rather than outcomes. Features launched, tools deployed, seats purchased. Rich Veldran, the CEO of GoTo, was direct about what that means: "If it's not changing how work gets done daily, it's just theater." GoTo has embedded AI into customer service workflows in ways that reduce call volumes, route issues faster, and free staff for higher-complexity work. The distinction Veldran draws is between AI layered on top of existing processes and AI that actually restructures how tasks are completed. The former looks like transformation; the latter is it.
The academic picture reinforces the same tension from a different angle. A National Bureau of Economic Research working paper analysing firm-level AI adoption found that companies investing in AI do tend to grow faster and command higher valuations, but aggregate productivity gains "remain elusive" in economy-wide data. The gap between what researchers observe in narrow task-level settings and what shows up in firm-level aggregates points to something important: AI is demonstrably effective at discrete tasks, but the organisational changes required to capture those gains at scale are slow, difficult, and often not happening. The paper notes that "adjustment costs and organisational complements are important," which is a precise way of saying that buying the tools is the easy part.
Part of what is driving the gap between activity and outcomes is pressure to be seen doing something. An AI consultant quoted in a St. Louis business survey put it in terms that are useful because they are blunt: "For a long time, investors rewarded companies for hiring people. We've exchanged one lazy proxy for another, and that is laying off people and blaming AI." Companies feel existential pressure to demonstrate AI credentials. Announcing headcount reductions attributed to AI efficiency, or announcing AI tool deployments across thousands of employees, signals seriousness without requiring proof that the underlying work is actually improving. It is easier to count seats than to measure whether outcomes changed.
The organisations that are making it work share a common approach: they redesign workflows around what AI makes possible rather than adding AI to existing workflows. Sixty-nine percent of workers in the Atlassian survey say their processes and workflows are not optimised for AI. That figure understates the challenge. It suggests the primary constraint on AI value in most organisations is not the capability of the tools but the willingness to do the slower, harder work of organisational redesign. Atlassian's conclusion is that the next wave of AI value will come not from adding more tools but from better coordination at the team level, treating AI adoption as an organisational change problem rather than a technology procurement one.
The pressure to prove ROI is intensifying. A significant share of IT leaders, in GoTo's own research, acknowledge that their organisations are not effectively measuring the return on AI investment. That situation is unlikely to persist as AI spending climbs. At some point, the boards and investors asking for the productivity numbers will require answers more specific than "we deployed AI to 50,000 employees." The companies that have been measuring outcomes all along, rather than inputs, will have the data. The ones that have been counting tools will face an awkward reckoning.