Read this article on our Substack The Impossible Brief
Efficiency is a false idol
How do we solve the human unknown in the AI equation? by Eleonore Pratoussy, Cassette Group Head of Marketing
Whilst businesses are scrambling to implement and adopt AI – obsessing over efficiency; we are forgetting to plan for the unbridled human and societal impact about to hit us at full speed.
After spending a few days in Cornwall at a conference, I can see that the current conversation is binary: we either talk about the tools and their use, or the risks to society inherent to AI. But no one is addressing how the two combine, how implementation decisions today directly shape whether we build human capacity alongside automation or damage it in the name of efficiency.
The future of work has less to do with the AI stack, and more to do with how we preserve intellectual rigour and critical thinking, how we teach it to our children (as AI is to them what Google Search was to us), how we manage and coach our juniors today so they are able to lead us tomorrow – lest we sacrifice them to AI agents in the first place.
Millions of people are experiencing this without real consideration of whether that transformation supports or undermines their ability to think, critique, and lead. This is a choice we are making, even if we are not admitting to it yet.
The Lost Generation
The data on AI deployment is impressive. Deloitte’s 2026 State of AI in the Enterprise survey of 3,235 global leaders found that 66% of organisations report productivity and efficiency gains as the primary benefit of AI adoption.[1]
There is limited research available on the impact of this rise in efficiency on people, but some interesting points from HBR and Anthropic:
A UC Berkeley study published in Harvard Business Review found that generative AI intensified work rather than reducing it. Employees worked faster and extended their work hours, leading to higher intensity and potential burnout rather than significant time savings. [2]
Anthropic’s randomised controlled trial with junior developers found that those using AI assistance scored 17% lower on comprehension tests than those coding manually.[3] When researchers examined 52 junior engineers more closely, a pattern emerged: developers who used AI for conceptual questions scored 65% or higher, whilst those who delegated code generation to AI scored below 40%.[3]
The productivity gains researchers expected to see failed to reach statistical significance. What did reach significance was the capability erosion.
When Deloitte asked what the biggest barrier to AI integration was, organisations pointed to the skills gap. When asked how they addressed it, the number one answer was education – teaching people to use the tools.[1] Not role redesign. Not workflow transformation for human capacity. Simply: that’s how you prompt the model, connect your artifact, vibe code that idea.
This pattern is not an AI problem, it is a design gap – the consequence of designing for speed without designing for human capacity. And it is avoidable.
Future Proofing for Human Transformation
The organisations that will capture long-term productivity advances from AI are not the ones using better tools but the ones redesigning work to build human capability alongside automation.
When we optimise for short-term productivity without considering human capacity, we create dependency. People cannot work without the AI because we never designed workflows that build their judgment alongside automation. We erode autonomy. The AI makes decisions; humans execute. We compound burnout, lack purpose. And we risk losing an entire generation of early AI-adopters to the god of short-term efficiency if we do not shift our lens.
Deloitte’s 2025 Global Human Capital Trends report shows that organisations investing in workforce development are 1.8 times more likely to report better financial results than those focused solely on technology implementation.[4]
The redesign is not about efficiency. It is about creating work structures where AI handles repetitive, low-autonomy tasks whilst humans gain time for judgment, relationship, and creativity.
This requires bilateral design: technical teams building systems that augment human judgment rather than replace it, whilst leadership redesigns roles to expand autonomy alongside efficiency.
The questions change: not ‘how much time can AI save?’ but ‘what judgment capabilities do we want humans to develop?’ Not ‘can AI do this task?’ but ‘where should humans retain decision authority even if AI could decide?’ The design is intentional when human capacity-building is measured alongside productivity gains and not assumed to follow automatically.
Most organisations are measuring what is easy to measure and assuming the human side will sort itself out - just like we once said the market would.
What We’re Trading
Every generation passes down what it means to be human – not just knowledge, but agency. It’s the first time in human history that a tool can think for us, rather than with us. The question therefore becomes: are we aware of what we are trading for efficiency, and do we choose to preserve human autonomy or encode human dependency?
We are making decisions right now about what it means to be a working human – what constitutes meaningful contribution, what we preserve and what we allow to erode. These define what kind of life we believe is worth living, what kind of work is worth doing, and what we are willing to sacrifice for short-term gain.
We're building this. The question is whether we're building it on purpose.
References
[1] Deloitte. (2026). The State of AI in the Enterprise 2026: The Untapped Edge. Deloitte AI Institute. Survey of 3,235 business and IT leaders across 24 countries, August-September 2025.
[2] Ranganathan, A., & Ye, X. M. (2026). “AI Doesn’t Reduce Work—It Intensifies It.” Harvard Business Review, February 9, 2026. Based on eight-month ethnographic study at 200-person tech company.
[3] Anthropic. (2026). “AI Assistance and Coding Skills.” Anthropic Research, February 2026. Randomised controlled trial examining how AI coding assistants affect skill development when learning new tools. Available at: https://www.anthropic.com/research/AI-assistance-coding-skills
[4] Deloitte. (2025). 2025 Global Human Capital Trends: The Worker-Employer Relationship Reimagined. Deloitte Insights.