David Wilkes is the CEO of Celcat, a Volaris Group-owned business, and he frequently writes about design-led thinking for the AI-native business leader. This article was adapted for Acquired Knowledge and originally appeared in David’s newsletter, Human + Machine | Design.

A 2026 Harvard Business Review article argues that AI does not reduce work; it intensifies it. Tasks expand, expectations rise, and what was once impressive becomes normal. What was once optional becomes required.

At the same time, Mental Health UK reports rising anxiety among young adults in the workplace, with two in five taking time off for mental health reasons.

Add to this the often-discussed topics of distraction, phone use, and declining conscientiousness. It is tempting to see these as separate issues. They are not. The pressures visible in younger cohorts may be an early signal of what the broader workforce will experience as AI adoption scales.

AI compresses cycles. We draft faster, analyse faster, and produce more variants in less time. The surface narrative is efficiency. Underneath, however, the number of parallel threads increases and the volume of inputs multiplies. Switching between tasks becomes constant. Context switching carries cognitive cost. Every interruption leaves residue and deep thinking fragments. Attention becomes shallow but continuous. If this pattern sustains, it does not create surplus capacity; it creates sustained cognitive intensity. Acceleration without redesign produces exhaustion.

I see this dynamic in my own workflow. I routinely operate alongside multiple AI agents across legal, marketing, HR, finance and thought leadership. Each agent accelerates output and extends capability. I can review contracts, shape campaign messaging, stress test policy, model financial scenarios and draft strategic pieces within the same hour. The leverage is extraordinary. Yet the cognitive demand is equally real. Managing parallel advisory streams requires deliberate sequencing, clear intent and disciplined stopping points. Without that structure, the day becomes a blur of intelligent prompts and partial completions. The lesson for me has been simple: multi-agent productivity only works when I design for focus first and layer capability second.

The leverage is extraordinary. Yet the cognitive demand is equally real.

There is also a psychological dimension. Many younger workers have grown up expecting disruption. They know technology reshapes industries quickly. Yet research suggests they report lower perceived control over their lives than previous generations. High expectations paired with lower agency produces strain. AI can recreate that pattern across all age groups. When systems suggest, optimise and allocate work, employees may feel less like authors of change and more like operators within it. If people comply with tools but do not reshape the system itself, agency contracts. Expectation without agency breeds anxiety. Technology does not determine whether people feel empowered. Organisational architecture does.

Conscientiousness adds another layer. It remains one of the strongest predictors of job performance, underpinning reliability, discipline and follow-through. Yet sustained conscientiousness requires uninterrupted attention. If work becomes a constant stream of prompts, meetings, dashboards and alerts, maintaining deep follow-through becomes harder for everyone. It is easy to describe declining conscientiousness as generational. It may instead be structural. When environments reward rapid switching and constant responsiveness, even disciplined individuals struggle to sustain depth. If AI intensifies these patterns, conscientiousness may erode not because people care less, but because architecture makes it harder to care well.

History shows that most large transformations fail not because of poor tools, but because organisations attempt them in the margins of “business as usual.” Teams remain overloaded, meetings proliferate and new initiatives layer on top of existing responsibilities. AI is particularly vulnerable to this failure mode. If it is introduced as an additional capability without freeing space, it intensifies busyness rather than replacing it. Successful transformation requires clarity of intent, protected time for focused experimentation, and the willingness to free top talent from part of their day job. It requires a clear North Star for a defined period and the discipline to embed reflection as a regular practice rather than an occasional luxury. Focus is not accidental; it must be designed. Without protected focus, acceleration simply compresses stress.

The structure of teams matters as well. Complexity scales non-linearly, and as groups grow larger, coordination overhead expands faster than output. AI projects demand both exploration and discipline. They benefit from compact teams with end-to-end ownership, diverse cognitive styles, and explicit authority. When too many people are involved, the noise of coordination overwhelms the signal of progress. When mandates are vague, teams default to incrementalism. If AI is to augment rather than exhaust, leaders must shape environments deliberately rather than assume benefits will emerge automatically.

AI also touches something deeper: professional identity. When a system performs part of your role faster or more accurately, it can feel like erosion of value. If unmanaged, this produces quiet disengagement or defensive behaviour. If addressed openly, it can become an opportunity to redefine contribution. The difference lies in whether employees see themselves as custodians preserving a legacy process or builders shaping a new system. That shift requires psychological safety, structured learning, and visible human authority in decision loops. Without those conditions, expectation outpaces agency, and anxiety follows.

Technology amplifies what already exists. If organisations reward speed over depth, AI will intensify fragmentation. If they reward focus, agency, and disciplined execution, AI will magnify those instead.

The real risk, then, is not that AI will eliminate work. It is that it will eliminate the natural boundaries that once protected cognitive energy. More tasks will be completed per hour. More responsiveness will be expected. More performance will be measured in real time. If this is layered onto existing cultures of busyness, we will not create a more capable workforce; we will create a more fatigued one. Young adult anxiety may be an early warning, not an isolated issue.

Technology amplifies what already exists. If organisations reward speed over depth, AI will intensify fragmentation. If they reward focus, agency, and disciplined execution, AI will magnify those instead. Leaders therefore face a design choice. They can protect deep work blocks, establish explicit boundaries around responsiveness, and integrate AI directly into workflows rather than bolting it on. They can build AI self-efficacy through structured learning and encourage reflection on how work is done, not just what is produced. They can frame AI as orchestration that supports human judgement rather than replacement that diminishes it.

Acceleration without architecture produces burnout. Acceleration with architecture produces transformation. The question is not whether AI will increase pace; it will. The question is whether we design systems where that pace is sustainable, where people feel agency within it and where conscientiousness is strengthened rather than eroded. That is not a technology problem. It is a leadership one.