Twenty-four pence of every pound spent on AI goes on managing complexity before any business value is realised, according to a survey of 12,021 IT decision makers across six countries by Freshworks.
The Global Cost of Complexity Report, released today, puts the UK's annual AI waste figure at an estimated £11.7 billion and describes the underlying dynamic as a "complexity tax" — overhead from integration friction, tool sprawl, and rework that consumes budget before the intended productivity gains materialise.
The research is focused specifically on mid-market organisations, defined as those with 250 to 5,000 employees. The argument is that this segment absorbs the cost of AI complexity differently from large enterprises: without the headcount to layer on dedicated oversight teams, fixing bad AI outputs and governing tool sprawl falls on the same teams that are also supposed to be benefiting from the technology.
The companies that move from purchase to performance fastest will turn AI from a complexity tax into a competitive advantage.
The numbers support that tension. Eighty-six percent of mid-market IT leaders say managing AI complexity has increased their team's workload. Eighty percent report that AI outputs introduce noise, errors, or rework — what the report terms "AI slop." At the same time, 89 percent plan to increase AI investment over the next 12 to 24 months.
The deployment timeline mismatch is a separate problem. Seventy-two percent of mid-market executives expect ROI within eight months of AI investment. Fifty-five percent of organisations say deployment alone takes between six and twelve months before meaningful ROI can even begin. Programs are being evaluated before they have had time to produce results.
Middle market businesses tend not to be early innovators and often lag in realizing full-scale implementation benefits until they are confident of ROI. Until then, smaller pilots and tests are often used to prove feasibility,
Tool fragmentation is compounding both issues. Mid-market organisations use an average of 4.2 AI tools, with 10 percent running seven or more. Only 33 percent have a formal, consistently applied AI governance framework. The buying response is becoming visible: 90 percent of respondents say they favour AI with built-in workflows over heavy configuration, and 54 percent are now buying AI capabilities rather than building in-house.
Fieldwork for the survey ran in March 2026, covering organisations in the US, UK, Germany, France, Singapore, and India.
