Sound-proof booths are as important as AI for your office
Your employees’ ability to concentrate is the biggest limiting factor on productivity, says Paul Armstrong
Corporate leaders are pouring billions into artificial intelligence in pursuit of a long-promised productivity boom. Early evidence is suggesting the gains may stall for a surprisingly familiar reason: human attention.
Research into what analysts now describe as ‘AI brain fry’ shows productivity rises when employees use one or two AI tools and peaks with three before declining once workers must supervise four or more systems. The Boston Consulting Group study examined 1,488 employees across large companies and found heavier AI oversight increased mental effort by 14 per cent, mental fatigue by 12 per cent and information overload by 19 per cent. Few bosses want to see data like this, as many organisations are deploying as many AI systems as possible.
Executives increasingly expect knowledge workers to operate as supervisors of machines. Analysts generate reports with one AI system, verify outputs with another, automate workflows through a third and combine results inside dashboards or document tools. Machines can produce unlimited information, but human attention still governs how much of that information can be processed. Big and small businesses have a problem brewing.
The open-plan office lesson businesses ignored
Corporate history offers a revealing precedent. Open-plan offices emerged from a straightforward management logic. Removing walls would increase collaboration, accelerate communication and allow companies to house more employees per square metre. Property costs fell while managers assumed productivity would rise. Evidence quickly contradicted the theory. Research reviewed in open offices shows employees consistently report lower concentration and higher distraction in open environments and that was without your coworkers shouting into Claude and co to get the latest numbers out.
Global workplace data reinforces the pattern. Noise and lack of privacy consistently rank among the largest barriers to productivity per Leesman Index data. Only a minority of workers report their workplace effectively supports focused work. Management redesigned offices around density and utilisation metrics rather than cognitive performance. Efficiency looked convincing on spreadsheets while concentration quietly collapsed on the ground. Leaders might need to prioritise soundproofing the offices if they don’t want a bigger problem on their hands when those offices actually fill up.
The digital version of the same mistake
Artificial intelligence risks repeating that error in digital form. Leaders deploy new tools in pursuit of efficiency without redesigning how work actually happens. Employees now operate inside open-plan offices filled with constant background noise while simultaneously managing messaging platforms, project dashboards and multiple AI systems generating additional information streams. Technology introduced to reduce effort ends up multiplying the number of decisions workers actually make.
Corporate behaviour often accelerates the problem. Organisations facing complexity typically deploy more systems rather than fewer. Each additional tool generates alerts, outputs and verification tasks that employees must interpret. AI copilots draft reports; workers must review them for accuracy. Automation tools trigger processes that require human confirmation. Digital infrastructure has expanded while the human capacity to process information remains fixed. Organisations discovering the limits of AI oversight will face rising burnout, higher turnover and expensive mistakes. Worse still, unchecked AI output can escape into the real world. Professional services firm Deloitte has already learned how expensive that mistake can become once clients start asking for refunds let alone major reputational damage.
The economic cost of that overload rarely appears on balance sheets. Studies of workplace interruption show knowledge workers switching tasks roughly every three minutes. Regaining focus after interruption can take far longer. Multiply those recovery periods across thousands of employees and entire organisations quietly lose weeks of productive time every year. Productivity does not collapse dramatically – it erodes gradually through fragmented attention.
Attention is becoming the real constraint
A deeper irony sits inside the productivity debate. Companies routinely allocate large budgets to artificial intelligence while ignoring far cheaper interventions that would improve concentration immediately. Acoustic design remains neglected across many corporate offices despite clear evidence linking noise to distraction. Acoustic ceiling panels, wall absorption materials, carpeting and sound-dampening partitions significantly reduce ambient noise. Trading floors and broadcast studios treat such infrastructure as essential because mistakes carry immediate cost. Corporate offices often treat it as cosmetic.
Many offices provide abundant meeting rooms and collaboration zones yet few spaces designed for uninterrupted work. Focus booths, single-person rooms and designated quiet zones allow employees to escape constant background noise and not annoy others. Attention is becoming the defining constraint in the AI economy. Machines can produce limitless analysis, recommendations, code and documentation. Human cognition remains responsible for judging accuracy, interpreting results and deciding what actions follow. Once machine-generated output exceeds the capacity of people to evaluate it, productivity gains stall regardless of how powerful the underlying technology becomes.Looking for a strategic advantage? Call a decorator at the same time as the AI expert. Companies that simplify tool stacks, limit how many systems employees supervise, create (and protect) concentrated work environments will extract greater value from artificial intelligence than competitors deploying AI with no strategy and focus on human attention. Firms capable of protecting it will make better decisions, move faster and waste less time interpreting the noise created by their own systems.