Why AI Productivity Depends on Human Imagination

Since the release of ChatGPT4 in 2022, the most vexing question in corporate boardrooms has been how to integrate AI technology into business processes and products, but this framing is itself part of the problem. A growing body of research shows that productivity gains come not from automating existing operations with AI, but from radically redesigning systems around this new technology—or even in collaboration with it. The limited success of AI adoption efforts is not driven by technological limitations, but by the self-imposed psychological boundaries of those tasked with adopting it.

In my book, The Technology Illusion: Dispelling the Myth of Digital Transformation, I explore how executives often fail to leverage the productive potential of digital technologies by refusing to fundamentally reimagine their products, services, and internal systems. Instead of adapting their corporate strategy to reflect the possibilities of a digital world, they integrate voguish technologies into their operations in ways that are flashy but essentially superficial. Rather than profoundly disrupting their own administrative structures, value-creation processes, and business models by taking the kind of risks necessary to reap the full productivity benefits of a transformative technology, cautious executives merely aim to put a reputational gloss on existing products and legacy systems. The result is a “technology illusion”, as executives focus on creating a myth of digital transformation while avoiding the sweeping changes necessary to make it a reality.

A reluctance to embrace radical change continues to bedevil corporate efforts to fully utilise digital technologies, and the advent of AI has dramatically increased the stakes. Corporations are spending billions on AI technology, yet an oft-cited study by MIT recently found that 95% of AI pilots generate no meaningful benefits, and a growing body of analytical work reveals why. The leading explanations fall into three main categories:

• • Experimentation over transformation: Writing in Harvard Business Review, Nathan Furr and Andrew Shipilov argue that “leaders are repeating the mistakes of the digital transformation by funding scattered pilots that don’t connect to real business value.” Under pressure to demonstrate their competence in a cutting-edge technology, executives attempt to pioneer applications themselves, when what they should be doing is assigning this task to agile teams of dedicated specialists. In Forbes, Andrea Hill reaches a similar conclusion, noting that “too many executives are green-lighting projects not because they solve a defined business problem, but because they feel they all need an AI initiative.” While it manifests in the proliferation of failed pilots, the underlying incentive to adopt AI for its own sake stems from the hype surrounding the technology, which also poses its own distinct challenge.

• • The hype bubble: When a new technology emerges, expectations can run far ahead of applications, as firms and investors imagine a future of business that is either slow to materialise (e-commerce) or proves largely fanciful (the metaverse). In a recent article for Fast Company, Faisal Hoque identifies an “AI hype bubble” as one of three discrete bubbles created by the technology. This analysis dovetails with the well-established (though not uncontroversial) concept of the “Gartner hype cycle,” in which an innovation is assumed to be transformative and then treated (and valued) accordingly, only to struggle through a “trough of disillusionment” when it fails to meet expectations. Some technologies never recover, but many do ultimately revolutionise productivity—just not on the timescale that their initial boosters envisioned. In the meantime, the hype surrounding the new technology often distracts executives from pragmatic applications and acute problem-solving. Instead of drilling down on how to use the technology effectively, they focus on advertising their use of it in an effort to ride the hype cycle.

• • Automating inefficiency: Assigning a process to a machine yields no productivity gain if the process itself is fundamentally flawed or flatly unnecessary. In Fast Company, Balkrishan Kalra identifies “process debt” as a major constraint on AI-driven efficiency gains. Drawing an analogy to “tech debt”, the long-term cost of embedded technologies that obsolesce over time, Kalra defines process debt as the sum of the “manual workarounds, inconsistent data practices, and inefficient workflows” that gradually accumulate in every organisation. Tasking AI with automating a cobbled-together system based on imperfect data bakes those inefficiencies into AI operations, resulting in output errors that require human intervention to identify and address. Also in Fast Company, Tomas Chamorrow-Premuzic and Alexis Fink describe the related challenge of the so-called “BS economy,” the vast amount of work that cannot be automated efficiently because it doesn’t need to be done at all. They cite “long and useless meetings, excessive documentation, convoluted chains of approval, and performative busyness that drive up headcount and slow down progress without delivering any appreciable improvement in outcomes,” arguing for a “purpose revolution” to match the productivity revolution promised by AI.

All three explanations point to the same core problem: business AI isn’t increasing efficiency because it is being deployed by executives who are trying to avoid disruption rather than embracing it. The authors of the MIT study point to this underlying challenge, which Jason Snyder elaborates on in Forbes: “Without friction, GenAI is theatre. Smooth demos impress, but without governance, memory and workflow redesign, they deliver no value. The companies that succeed are those that engineer for friction, calibrating it rather than eliminating it.” Executives have plenty of reasons to avoid friction—professional investment in maintaining budgets and headcounts that could be reduced by AI, the daunting prospect of confronting bureaucratic inertia, or even the creeping suspicion that some of their own responsibilities may fall under the rubric of the BS economy. Acknowledging these incentives and destigmatising them are the first steps toward effecting the deep reforms necessary to unlock AI’s productive potential.

The fraught process of AI adoption reflects deeper challenges around technological transformation, and lessons from the flawed integration of other technologies into business processes can yield important lessons for AI uptake. The Institute of Information Systems and Digital Business at the University of St. Gallen (HSG), working in cooperation with Boyden Global Executive Search, recently studied 42 Swiss companies of various sizes and in different industries. The study found an incomplete and uneven process of digital transformation. Around 60% of the companies surveyed report that at least half of their workforce can use digital tools, including AI applications, confidently. While a basic level of capability has been achieved, few companies have cultivated the workforce skills profile necessary to make agentic AI or other complex technologies productive across the board. Some companies report clear advantages and a stronger market position due to digital transformation, while others perceive adopting new technologies as largely a burden, and a few have registered significant setbacks as expensive upgrading initiatives have failed to yield cost-effective results. Overall, the study finds that measurable benefits only become apparent when companies have established mature digital processes, successfully expanded their digital offerings, and strategically anchored digital transformation at the top management level.

The HSG study confirms the core challenge of AI adoption. Executives overwhelmingly focus their digital transformation efforts on immediately visible business and process effects while neglecting the deeper organisational restructuring and cultural anchoring of digital skills needed to make new technologies productive. As a result, the financial impact of digital transformation on the supply side has been modest. On average, digital offerings account for just 10% of the revenue of surveyed firms, and only a small fraction of firms generate more than half their revenue from the digital space. Digital offerings account for just under 30% of the average portfolio, and the automation rate for central processes is also around 30%. While digital transformation is driving value creation, even well-established technologies remain underutilised.

Business AI, with its disorienting array of potential use cases, obscure internal mechanics, counterintuitive weaknesses, and ever-evolving capabilities, presents challenges to executives that extend far beyond those of any previous digital technology. Merely adopting AI will not create efficiency or competitive advantage, especially if executives cling to a systems-oriented pre-AI mindset in which the technology is viewed as simply a new means to automate existing processes and augment offerings already on the market. Productivity gains emerge when organisational architectures, talent strategies, and value-creation models are rebuilt to harness the technology’s unique capabilities rather than shoehorning it into legacy structures shaped by a world of business that no longer exists. This shift demands a willingness to tolerate friction, experiment with new forms of governance and collaboration, and dismantle processes that serve only to perpetuate the status quo. Companies that embrace this deeper transformation will move beyond the illusion of digital progress and begin constructing business systems fitted to an AI-mediated economy. Those that do not will find themselves spending heavily on technologies that deliver only symbolic innovation and ultimately ceding advantage to more flexible and imaginative rivals.

Leveraging the full potential of AI will require executives to confront an uncomfortable truth: the threat to productivity is not the technology but the institutions that refuse to evolve alongside it. For executives who prioritise reputational signalling over deep structural change, allow managers to protect their organisational fiefdoms rather than disrupt the status quo, and fund isolated pilots while refusing to redesign core systems, AI will be yet another mirage in a long line of failed transformations. However, executives who are willing to challenge their own incentives and those of their colleagues, dismantle obsolete workflows, and embed digital strategy at the centre of organisational life will be positioned to catalyse the kind of productivity revolution that digital transformation has long promised but seldom delivered. The choice is not between adopting AI or falling behind. The choice is between redesigning enterprises to make AI meaningful or preserving a system that ensures it never will.

References
https://www.economist.com/finance-and-economics/2024/08/19/artificial-intelligence-is-losing-hype
https://hbr.org/2025/08/beware-the-ai-experimentation-trap
https://www.msn.com/en-us/money/personalfinance/the-silent-killer-of-ai-success/ar-AA1MlTgk
https://www.fastcompany.com/91400857/there-isnt-an-ai-bubble-there-are-three-ai-bu
https://www.fastcompany.com/91391631/how-ai-is-exposing-the-bs-economy-ai-work-productivity
https://mlq.ai/media/quarterly_decks/v0.1_State_of_AI_in_Business_2025_Report.pdf
https://www.forbes.com/sites/jasonsnyder/2025/08/26/mit-finds-95-of-genai-pilots-fail-because-companies-avoid-friction/
https://www.forbes.com/sites/andreahill/2025/08/21/why-95-of-ai-pilots-fail-and-what-business-leaders-should-do-instead/

Fragen, Feedback und Kommentare zu diesem Beitrag senden Sie bitte an kontakt@acent.de