As generative artificial intelligence becomes ubiquitous, organisations are rushing to eliminate friction from their workflows in the pursuit of hyper-efficiency. However, new research from behavioural scientist Dr. Chengwei Liu reveals that making AI too easy to use is a profound strategic risk. By removing the cognitive hurdles that force critical thinking, businesses fall into a "productivity trap"—trading long-term innovative capacity for short-term cost savings. To survive the AI era, forward-thinking leaders must introduce "strategic friction," purposefully designing cognitive speed bumps into their operations to preserve the absorptive capacity, divergent thinking, and talent moats necessary to outmanoeuvre the competition.
A mid-morning walk through the towering, biophilic concourses of CapitaSpring in Singapore’s Central Business District reveals a striking tableau of modern enterprise. Behind the glass walls of multinational headquarters and agile venture capital firms, screens flicker with the same hypnotic rhythm. Knowledge workers are generating market reports, debugging code, and drafting strategic communications at a pace that would have seemed supernatural just three years ago. With a few keystrokes, an artificial intelligence synthesises a competitive analysis that once required a week of laborious research. It is a vision of frictionless, hyper-efficient capitalism.
But beneath this veneer of peak productivity lies a quiet, insidious threat. In our race to automate the mundane, we are inadvertently outsourcing our capacity for original thought.
We have long operated on the assumption that faster is better, that frictionless workflows unlock human potential, and that AI is the ultimate cognitive lever. Yet, as the novelty of large language models settles into everyday utility, a counter-narrative is emerging from the vanguard of behavioural science. It suggests that our relentless pursuit of ease is hollowing out the very talent we need to drive future growth. If necessity is the mother of invention, what happens to human ingenuity when the necessity to think deeply is entirely engineered away?
The Productivity Trap: When "Good Enough" Becomes the Enemy of Great
Dr. Chengwei Liu, an Associate Professor of Strategy and Behavioural Science at Imperial College Business School and author of the forthcoming book The Smart Contrarian, has spent his career examining how luck, bias, and cognitive patterns shape high-stakes decision-making.
The mechanism behind this stifling effect is what Liu terms the "productivity trap".
However, the long-term consequences pose an existential threat to corporate competitiveness. When "good enough" solutions become free and universally accessible across an organisation, the behavioural incentives shift drastically. Liu’s research demonstrates that the rate at which employees reuse readily available information skyrockets, while the rate of independent exploration, trial-and-error, and experimentation plummets.
Why spend three days wrestling with a complex architectural problem when an AI can deliver a perfectly adequate framework in three seconds? The danger is that, over time, teams increasingly converge on a limited, homogenised number of approaches.
Rogers’ Paradox and the Economics of Cognitive Free-Riding
To understand why this happens, we must delve into the economics of knowledge production. In a recent paper published in Management Science, Liu and his colleagues explored the tension between fast learning and sustained exploration. Strategy theories have traditionally argued that "more is better"—that fast, frictionless sharing of information enhances collective performance. However, organisational learning theory warns of a "less is more" dynamic, cautioning that rapid learning causes premature convergence.
This brings us to a concept known as Rogers’ paradox. In any ecosystem, producing new knowledge is costly. It requires time, effort, and a high tolerance for failure. When knowledge diffuses too easily—as it does when an entire workforce is hooked up to a centralised generative AI platform—individuals are highly incentivised to free-ride on existing knowledge rather than produce their own.
If everyone in your marketing department uses the same foundational model to generate campaign ideas, they are all pulling from the same latent space of human knowledge. The friction of sharing is zero. But because no one is doing the costly work of exploring the fringes of the intellectual frontier, collective knowledge stagnates. The average payoff for the organisation is no greater than if everyone had just worked independently without the technology.
The Singapore Paradigm: Re-evaluating Hyper-Efficiency
There is perhaps no geography where this tension is more acutely felt than in Singapore. The city-state’s economic miracle is largely predicated on an obsession with efficiency, optimisation, and rapid technological adoption. From the algorithmic precision of the PSA port terminals to the rollout of the National AI Strategy 2.0 (NAIS 2.0), Singapore has built its reputation as the ultimate beta-tester for the future.
In the local corporate culture, heavily influenced by a pragmatic, kiasu (fear of losing out) mentality, the adoption of generative AI has been swift and unforgiving. Firms in one-north and Marina Bay are scrambling to integrate LLMs into their tech stacks, driven by the fear that competitors will outpace them in cost-efficiency.
But viewing AI purely through the lens of efficiency is a strategic misstep for a mature economy. Singapore is no longer competing on cheap labour or simple execution; it competes on premium, knowledge-based services, cutting-edge R&D, and geopolitical strategic advisory. These are domains that require profound originality and nuanced judgement.
If Singaporean enterprises fall into Liu’s productivity trap—using AI to simply execute the same tasks faster—they risk commoditising their own workforce. If a legal associate in Raffles Place and a legal associate in London are using the exact same AI model to draft a contract, the only differentiator left is the human capacity to identify edge cases, interpret cultural nuances, and strategise beyond the algorithmic mean. By attempting to optimise every minute of the workday, Singaporean firms risk breeding a generation of "prompt managers" whose capacity for deep, independent, unassisted thought has withered. The execution premium vanishes, and with it, the global competitive edge.
The Antidote: Designing "Strategic Friction"
If banning or severely limiting AI is not an option—doing so would invite catastrophic opportunity costs in short-term savings—how do business leaders balance the efficiency of AI with the preservation of creative thinking?
The solution, according to Liu, lies in the deliberate introduction of "strategic friction".
Strategic friction involves intentionally adding small, calibrated hurdles to the individual use of AI to raise the total level of knowledge at the organisational level.
This works through a mechanism known as "absorptive capacity"—the necessity of prior investment in one’s own knowledge before one can effectively learn from, or evaluate, others.
Think of it as cognitive weightlifting. You cannot build muscle by watching someone else lift weights, nor can you build absorptive capacity by having an AI instantly solve your problems. You must experience the resistance. Absorptive capacity forces employees to grow their understanding, allowing them to evaluate, adapt, critique, and improve upon an AI’s output, rather than blindly copying it.
When strategic friction is applied, pure free-riding is discouraged.
What Strategic Friction Looks Like in Practice
Crucially, strategic friction is not about creating bureaucratic red tape or managerial bottlenecks just for the sake of it.
Here is how forward-thinking organisations can weave strategic friction into their operations:
1. The "Blank Page" Mandate for Strategy Formulation
Before an executive or creative team is permitted to use AI to generate options for a new product launch or a strategic pivot, they must first spend a designated period mapping out their own hypotheses on a blank page. By forcing the human brain to retrieve information, connect disparate dots, and face the discomfort of the unknown, the team builds the absorptive capacity required to later evaluate the AI’s suggestions critically. They use the AI to stress-test their ideas, rather than relying on the AI to generate the ideas in the first place.
2. The Mandatory "Devil’s Advocate" Protocol
If an AI provides a "good enough" solution to a complex logistical problem, strategic friction dictates that the solution cannot be accepted at face value. A designated team member must be tasked with explicitly dismantling the AI’s logic, identifying its blind spots, and proposing an opposing viewpoint. This friction forces the team to understand the why behind the solution, ensuring that they do not lose their grip on the underlying mechanics of their own business.
3. Bifurcating the Workflow: Routine vs. Complex
Leaders must carefully calibrate where friction belongs. Not every task requires it.
Building a Talent Moat for the Contrarian Enterprise
In his upcoming book, Dr. Liu explores the concept of the "smart contrarian"—the person who knows exactly when the crowd is wrong and what to do about it.
The firms that ultimately win with AI will not be those that achieve the highest degree of automation. Automation is a commodity; everyone will have it. The winners will be those who use technology to handle the mundane while rigorously protecting their workforce’s ability to engage in the deeply human work of judgement under uncertainty.
This requires a profound shift in how we evaluate employee performance. If we reward knowledge workers solely on their speed and volume of output, we are incentivising them to lean entirely on AI, thereby accelerating cognitive atrophy. Instead, leaders must begin rewarding the quality of questions asked, the depth of critical evaluation, and the willingness to pursue divergent, non-consensus ideas.
Singapore’s trajectory as a global hub hinges on its ability to navigate this transition. Its institutions, both public and private, must pivot from celebrating sheer operational efficiency to cultivating deep intellectual resilience. A Smart Nation cannot be populated by citizens who have outsourced their critical faculties to a server farm in California.
The most dangerous thing an organisation can do with AI is make it perfectly, seamlessly easy to use.
Key Practical Takeaways
- Audit Your AI Use Cases: Map out where AI is currently being used in your organisation. Divide these tasks into "Routine/Execution" (where zero friction is desired) and "Complex/Strategic" (where friction is necessary).
- Enforce 'Think First, Prompt Second': Implement rules that require teams to establish their own hypotheses, frameworks, or first drafts independently before running them through an LLM for refinement.
- Reward Divergent Thinking: Adjust KPIs to ensure you are not inadvertently rewarding employees purely for the speed of AI-generated output. Reward rigorous critique, novel problem-solving, and the uncovering of algorithmic blind spots.
- Cultivate Absorptive Capacity: Ensure your training programmes focus on deep domain expertise rather than just "prompt engineering." Employees cannot accurately evaluate an AI's output if they do not possess the underlying knowledge to spot its flaws.
- Embrace the Discomfort of Friction: Educate stakeholders that a slightly slower, more deliberate process in strategic planning is not a sign of inefficiency, but a necessary investment in long-term innovation and talent retention.
Frequently Asked Questions
What is the "productivity trap" in the context of AI?
The productivity trap occurs when the ease of using AI for instant, "good enough" solutions causes a workforce to stop exploring, experimenting, and thinking critically. While it boosts short-term efficiency and lowers costs, it slowly erodes the organisation's capacity for genuine innovation and independent problem-solving.
What does Dr. Chengwei Liu mean by "strategic friction"?
Strategic friction is the deliberate introduction of cognitive hurdles or manual steps into a workflow before AI can be utilised. Rather than adding mindless bureaucracy, it forces employees to invest mental effort to understand and evaluate information, thereby preserving their learning, critical judgement, and creativity.
How does "absorptive capacity" protect a business from AI homogenisation?
Absorptive capacity is the ability to build expertise by stacking new learning onto existing knowledge. By forcing employees to wrestle with problems rather than just copying AI outputs, they maintain the deep contextual understanding required to adapt, critique, and improve upon ideas, ensuring the business retains its unique competitive edge.
Further Reading & Resources:
No comments:
Post a Comment