Boston Consulting Group surveyed 1,488 full-time US workers and found a threshold that most AI deployment strategies are built to cross. Productivity increases when employees use one to three AI tools. At four or more, it collapses. Workers beyond the threshold report 33 per cent more decision fatigue, 39 per cent more major errors and a nine-percentage-point rise in intent to quit.
BCG calls it "AI brain fry." The label is vivid. The operating model implication is more precise: organisations are generating measurable cognitive harm by deploying AI tools into workflows that were designed before those tools existed.
The standard response to this finding will be tool rationalisation: reduce the number of tools, consolidate vendors, tighten procurement. That response treats the symptom. The problem is architectural.
Each AI tool deployed into an existing workflow adds a supervision layer. Someone must decide when to use the tool, how to evaluate its output, when to override it and how to reconcile its recommendations with those of other tools operating in the same workflow. That supervision burden falls on the individual worker. It is internal to the task, surfacing only as slower decisions, higher error rates and attrition.
The structural failure is that the workflow was designed for a human doing the work alone. AI tools were added without redesigning the workflow around a different division of cognitive labour. The tools accelerate individual steps. The workflow absorbs those gains as coordination overhead. At three tools, the overhead is manageable. At four, the supervision burden exceeds the productivity gain. The result is a net loss disguised as an AI investment.
The BCG data makes one structural implication unavoidable for any leadership team deploying AI across multiple functions. Every workflow that uses AI tools requires a deliberate redesign for human-AI collaboration. The workflows that were designed for humans alone and received AI tools on top are the workflows producing the cognitive harm BCG measured.
Workflow redesign means deciding which cognitive tasks belong to the AI, which belong to the human and where the handoffs sit. It means designing the supervision points explicitly, so every worker operates within a defined structure. It means treating tool governance as a workflow design discipline, owned by the people who design how work gets done.
Organisations that have already deployed four or more AI tools into unredesigned workflows are likely experiencing the effects BCG identified. The remedy is the same as the prevention: redesign the workflow before adding the next tool. The 33 per cent decision fatigue figure is the cost of skipping that step.