AI for HR: Next up – performance reviews
CHROs must lead AI adoption in HR, from performance reviews to strategy, balancing risks and rewards while enhancing organizational impact....
Artificial Intelligence
Is your AI strategy built for a world that no longer exists?
To truly transform, organizations must overcome strategic and conceptual pitfalls. Here is why so many leaders continue to get it wrong.
The true promise of artificial intelligence (AI) has always been to make the impossible possible – not just to accelerate our current work, but to invent entirely new ways of creating value. It is a catalyst for reimagining the very foundations of business. This creates a challenge for leaders to avoid the gravitational pull of incrementalism – the temptation to apply this revolutionary technology to simply optimize the past, when a total transformation is about to occur.
Unlocking that transformation requires navigating two distinct pitfalls. The first is strategic: understanding why today’s most common AI adoption strategies are built to deliver marginal gains, not breakthrough value. The second is conceptual: fully grasping the exponential nature of the underlying technology, a force poised to make our best linear plans obsolete. By confronting these challenges, we can begin to forge a new mental model designed not just to implement AI but to build a truly AI-native enterprise.
The three paths of AI adoption today
At InferenceCloud.ai, our perspective comes from the sharp end of change. For the past three years, our work has taken us into the trenches with dozens of organizations across the diverse business and regulatory landscapes of Europe, Asia, and Latin America. We’ve implemented AI solutions in demanding sectors ranging from financial services and pharmaceuticals to industrial and automotive. This experience is forged from navigating both failures and successes that have saved clients significant sums and generated millions in new revenue. These on-the-ground observations have revealed two dominant, yet flawed, strategic postures, highlighting the need for a third, more pragmatic path.
Type A: Bottom-up adoption is the default path, predicated on the idea that access equals integration. Leadership distributes AI tools to the workforce, under the assumption that productivity will naturally follow. Success is measured not by its impact on core KPIs, but through proxy metrics like adoption rates and feel-good surveys. Despite feeling manageable, it is structurally incapable of success due to two fundamental barriers:
- The value capture gap: an employee using an AI tool to write an email 15% faster is a real benefit to the individual, but it is a productivity gain that dissipates into organizational slack; it does not reduce headcount or increase revenue.
- The human bottleneck: an AI-assisted process is introduced to replace a human one; employees are incentivized to protect their own value.
Type B: Top-down revolution is a rarer but more potent approach. It views AI not as a tool for workers, but as a new foundation for the work itself. The starting point is a first-principles look at the business, ruthlessly identifying opportunities to re-architect entire processes around AI. A Type B organization bypasses the human bottleneck entirely by making clear-eyed business decisions, not emotional ones. It will happily and rationally accept 90% of the previous quality if it comes with a 50% reduction in cost or time. These savings can be subsequently invested in other ways of improving ultimate quality and value to provide lasting competitive advantage. However, while Type B represents the theoretical ideal for disruption, it requires a level of political will, risk tolerance, and cultural upheaval that most large companies are not built to withstand.
This tension between the impotence of Type A and the impracticality of Type B is why a third, more pragmatic path emerges.
Type C: Strategic hybrid is the path of smart, persistent evolution. These firms identify a single, high-leverage problem – a workflow defined by extreme cost or inefficiency – and deploy a dedicated team to solve it with a clear, metric-driven goal. By demanding results against existing KPIs, it delivers tangible organizational value. This accumulates internal credibility, builds the necessary skills, and provides organizational air cover for the next, more ambitious project. It is the only approach that allows an established organization to earn the right to execute a true transformation over time.
Given these dynamics, the Type C approach appears to be the most rational choice for today, yet it is dangerously incomplete.
“This explains why many analysts believe AI is falling short of its transformative potential, and why surveys show that firms are failing to create financial returns from AI investments.”
The horizon problem: Trapped in the wrong conversation
The flaw in even the best Type C thinking of persistent evolution is that it is linear in a world that has become undeniably exponential. Despite believing that they are implementing AI strategies correctly, leaders are often stuck in the optimization trap – obsessing over tactical questions that may soon be entirely irrelevant. This explains why many analysts believe AI is falling short of its transformative potential, and why surveys show that firms are failing to create financial returns from AI investments. In their haste to adopt, managers are often looking at a landscape of incremental improvements and mistaking it for the destination.
Consider the critical business process of customer lead generation.
- The linear trap asks, “How can we use AI to optimize our lead generation process?” This leads to projects that use AI to score existing leads, automate email follow-ups, or personalize website content – useful, incremental improvements that aim to make the existing human-driven funnel more efficient.
- The exponential leap comes from asking, “What happens when an AI sales development agent is able to autonomously identify ideal customer profiles, find prospects, conduct initial outreach, and deliver fully qualified, high-intent leads to our human sales team?” In this scenario, the entire top-of-funnel process is not just optimized; it is fully automated and reimagined. The debate over which tool provides better lead scores becomes a meaningless distraction when a system can generate the outcome on its own.
This dynamic scales from a single process to the entire enterprise. We have seen this movie before with Blockbuster, which focused on optimizing its physical stores while Netflix was rewriting the rules of media distribution. Consider the scale of that linear miscalculation. At its 2004 peak, Blockbuster was a global titan with 9,000 stores and nearly $6bn in revenue. Just four years prior, it had famously refused to acquire a struggling startup called Netflix for $50m. The collapse wasn’t a slow decline – it was a cliff. By 2010, as Blockbuster filed for bankruptcy under nearly $1bn in debt, Netflix’s subscriber base had surged past 20 million and it was well on its way to a valuation that would eventually eclipse $100bn.
The terrifying truth for leaders today is that AI is not a Netflix-level disruption for just one industry. It targets the universal substrates of knowledge work – how we think, communicate, and organize. It is Netflix-level disruption for every company, everywhere, all at once.
This isn’t speculation; it is an economic reality driving a ruthless talent war among the world’s top AI labs. These firms are not offering key researchers tens of millions of dollars in compensation to build a slightly better lead-scoring tool. The prize is total: a technology that provides a wholesale replacement of entire capabilities. While some remain trapped in a linear debate about assisting current processes, they are building the technology to make those processes irrelevant.
The danger is not that neither artificial general intelligence (AGI) nor artificial super-intelligence (ASI) will arrive overnight. It is what the Blockbuster story proves in stark financial terms: that the accelerating pace of progress, fueled by unprecedented investment, can make a multi-year transformation plan obsolete before it is even finished.
Upgrading your strategic operating system
To navigate this threat, we must avoid the fatal error of focusing on the wrong problem. Consider the advent of digital cameras. The entire photography industry, including its inventor, Kodak, spent years obsessing over a linear challenge: “How do we make it easier to print digital photos?” They were trying to fit a revolutionary technology into an outdated workflow.
This linear focus was rendered irrelevant by the smartphone and the rise of platforms like Twitter and Instagram. They didn’t attempt to solve the printing problem; they created an entirely new paradigm of frictionless digital sharing that made the need to print photos obsolete, capturing untold new value in the process. To avoid having your organization’s “Kodak moment” with AI, we must stop asking how AI can improve old processes and instead build a new framework to reimagine business functions from the ground up.
This requires considering your organization across four core components, replacing the linear questions of yesterday with the exponential-type questions of the future.
1. Process and organization
- Linear thinking: Optimizing existing human workflows and mapping out how AI tools can assist at each step.
- Exponential mindset: Viewing the organization as a data-flow architecture. The framing is not teams and workflows, but the optimal, automated architecture for the desired outcome. At which specific points do humans inject unique, non-automatable value, like judgment, negotiation, or empathy?
2. People
- Linear thinking: Focusing on tool fluency (“Will our people adopt this tool?”) and a narrow view of value capture (“Can we save two hours per employee?”).
- Exponential mindset: Assuming execution will be commoditized and automated. This creates the central leadership challenge of the next decade: “How do we move our workforce up the value chain from executors of tasks to architects of AI-driven systems?” This is not a re-skilling exercise; it is a profound re-conception of talent, where human value shifts to strategic intent, contextual awareness, and creative problem-framing.
3. Knowledge and data
- Linear thinking: Focusing on data readiness and integration for use in current AI tools.
- Exponential mindset: Assuming that generic intelligence will be a commoditized utility, like electricity. The only defensible asset is proprietary data that creates a unique competitive moat. The question becomes: “What data can we generate through our unique operations that no one else has, and how can we use it to create an intelligence advantage our competitors cannot replicate?”
4. Strategy and product
- Linear thinking: Asking, “How can we use AI to make our existing product 10% cheaper or add a new feature?”
- Exponential mindset: Asking, “How can unlimited, cheap intelligence fundamentally transform the value we provide?” For example, a company selling CRM software stops thinking about better contact management and starts thinking about delivering an autonomous sales development agent. A media monitoring company moves from providing dashboards to running predictive narrative simulations to find the optimal communications strategy before a campaign even launches.
A Type A organization, focused on tinkering, is building no muscle at all. It will be a spectator to its own demise.
Building adaptation muscle
Such questions can feel abstract, even overwhelming. They present a vision of a future so different from today’s reality that the path from here to there seems impossibly steep. How can a successful organization based on linear optimization possibly make such a leap?
This is where we must revisit the paradox of the Type C strategy. Viewed in isolation, these small, persistent projects seem merely linear – a safe, incremental path. But their true purpose is not just the immediate ROI. Their real value is in preparing the organization for the exponential leap to come.
This is why the Type C approach, while insufficient as an end state, is critical as a starting point. Those small, focused projects are not just about delivering a product; they are the gym where your organization builds its ‘adaptation muscle’ – the skills, culture, and data-first mindset required to compete in the next era. Each successful project is a training session, strengthening your ability to re-architect processes, re-evaluate talent, and build data moats.
A Type A organization, focused on tinkering, is building no muscle at all. It will be a spectator to its own demise.
A Type C organization is in training, preparing itself for a fundamentally different contest. The future will not be won by the company with the most AI tools, but by the one with the highest capacity to adapt. The workout starts now.
Authors
Related
2026 AI trends: What leaders need to know to stay competitive
Is your organization AI-ready? IMD professors and researchers predict 2026 trends in artificial intelligence. Read on and prepare to innovate – or be left behind. ...
How to get ahead in the AI cybersecurity arms race
AI is reshaping cybersecurity, arming both hackers and defenders. Learn how to stay ahead in the fast-evolving AI cybersecurity arms race....
From low-code to vibe code: Why you may no longer need software engineers
Vibe coding lets anyone build apps in plain English using AI, unlocking innovation and speed—but businesses must manage security, compliance, and quality risks....
Learn Brain Circuits
Join us for daily exercises focusing on issues from team building to developing an actionable sustainability plan to personal development. Go on - they only take five minutes.
Explore Leadership
What makes a great leader? Do you need charisma? How do you inspire your team? Our experts offer actionable insights through first-person narratives, behind-the-scenes interviews and The Help Desk.
Join Membership
Log in here to join in the conversation with the I by IMD community. Your subscription grants you access to the quarterly magazine plus daily articles, videos, podcasts and learning exercises.