MIT research shows most corporate AI initiatives fail, and many erode engagement and cognitive capabilities.

Approaching AI as an automation tool isn't returning value, and in many cases it's robbing workers of meaning.

There are three ways to take a more effective, human-centered approach: optimize, elevate, and innovate.

With intentional design, AI can unlock our highest human capabilities.

In early 2024, Klarna, the Swedish fintech company, announced that its AI-powered chatbot had assumed the workload equivalent of 700 customer service agents, managing millions of conversations across dozens of languages. The company reduced its workforce by roughly 40 percent. An IPO filing followed and, for a brief moment, Klarna appeared to be winning the AI-driven future of work (Haun, 2025).

But the story did not end there. As months passed, challenges emerged that efficiency metrics had obscured. Customers dealing with nuanced or emotionally charged situations encountered responses that felt formulaic and hollow. Satisfaction scores declined. The quality of service, particularly for complex cases requiring human empathy and judgment, deteriorated in ways that the gains in efficiency could not offset. By mid-2025, Klarna began bringing back human agents. CEO Sebastian Siemiatkowski acknowledged that the company had prioritized cost reduction over quality, and that approach proved unsustainable (Haun, 2025).

Klarna’s story is not unique. A growing body of evidence suggests that a similar tale is playing out, in quieter ways, across thousands of organizations and impacting millions of employees.

The Returns That Never Came

The gap between AI investment and AI results was one of the defining business stories of 2025. A major MIT study concluded that roughly 95 percent of enterprise generative AI pilots produced no measurable impact on profitability (Challapally et al., 2025). McKinsey’s QuantumBlack research unit arrived at a similar conclusion from a different angle. Their analysis found that while adoption of generative AI had become nearly universal, most companies reported no meaningful financial benefit (Sukharevsky et al., 2025). One analysis of 2025 enterprise data estimated that the average large company lost $7.2 million per failed AI initiative, with most abandoning more than two projects over the course of the year (Pertama Partners, 2026).

The financial shortfalls, however, seem to be masking a deeper problem. A growing body of research suggests that efficiency-focused AI deployment is changing the experience of work in ways that erode the very things that keep employees engaged, effective, and motivated.

In a series of experiments involving more than 3,500 participants, researchers found that working alongside generative AI on professional tasks like writing and brainstorming did improve output quality—but it came at a psychological cost. When participants transitioned to subsequent tasks without AI assistance, their intrinsic motivation dropped measurably, and their reported boredom increased (Liu et al., 2025). The researchers attributed this to a diminished sense of personal control. When AI handles the cognitive load, the human experience of authorship and agency fades, and with it the internal drive that sustains engagement over time.

In a longitudinal study at MIT, EEG scans revealed that those relying on AI exhibited progressively weaker neural connectivity compared to a control group working without technological assistance (Kosmyna et al., 2025). The participants using AI showed reduced brain engagement over time, with consequences for the long-term quality and originality of their independent work.

At the organizational level, the picture is equally sobering. An analysis of workplace data, drawing from more than 163,000 employees across over a thousand companies, found the proportion of employees showing signs of disengagement climbed 21 percent, reaching nearly one in four workers (ActivTrak, 2026). Time spent in AI tools grew eightfold, but sustained focus declined to a three-year low, and weekend work increased by more than 40 percent.

As efficiency-focused AI initiatives optimize workflows, some of the tasks that give employees a sense of meaning are taken away. Researchers at the University of Florida have even proposed a clinical framework for the psychological fallout, describing what they call AI Replacement Dysfunction: a pattern of anxiety, identity loss, and diminished self-worth tied to the experience of being made peripheral by technology in one’s own work (McNamara & Thornton, 2025).

These findings point to a fundamental problem with how most organizations are approaching AI. The dominant question—What can we automate?—treats work as a collection of tasks to be optimized.

But work is not just a set of tasks. It is also the context in which people form relationships, develop skills, feel a sense of contribution, and grow as professionals. When leaders focus exclusively on what AI can do instead of humans, they risk dismantling the scaffolding that supports the motivation, creativity, and resilience their organizations depend on.

This is the efficiency trap, and it explains why many AI investments are failing to produce returns. It is not that the technology doesn’t add value; it is that the lens through which it is being applied systematically overlooks the human conditions that make teams thrive.

Optimize, Elevate, Innovate

A better question is: What makes our people’s work meaningful, and how do we protect that while using AI to unlock something even better?

Through our consulting work with organizations navigating AI transformation, we have developed a framework designed to move leaders past the efficiency trap. We call it optimize, elevate, innovate, and its power lies in its human-centered approach to AI.

Optimize is the natural starting point and the most obvious application. It involves using AI to absorb the low-meaning, high-friction work that drains people’s energy without contributing to their sense of meaning. Things like routine data entry, repetitive processes, and standardized communication.

At HubSpot, the events team recently unveiled a bot they built to help any employee plan a sustainable company event—complete with budget guidance, location recommendations, and carbon footprint calculations. The tool freed the events team from repetitive planning requests so they could focus on more meaningful strategy, impact measurement, and advancing the company’s sustainability goals. At the same time, employees across the company were able to plan high-quality events that drove new business without waiting for central support.

This approach supports the practice of job crafting—the proactive effort to reshape one’s role to better align with personal strengths and values. Emerging research suggests that when AI adoption is handled thoughtfully, it actually stimulates employees to redesign their own work in more meaningful directions rather than passively accepting technological change (Xu & Qin, 2026).

Elevate is where true transformation begins. Rather than asking what AI can take over from people, elevation asks what AI can do to make people better at the work that matters most. Research from the Harvard Business School demonstrated that when management consultants used AI to support complex analytical tasks, the quality of their work improved substantially (Dell’Acqua et al., 2023). AI did not replace human judgment—it augmented it and raised the bar.

P&G recently ran an AI experiment across its commercial and R&D departments. They found that AI-augmented cross-functional teams produced breakthrough ideas at three times the rate of other teams. Technical specialists started proposing more commercially viable solutions, while commercial professionals developed more technically sound ideas. AI didn’t just speed up existing thinking; it helped people contribute meaningfully outside their core expertise, breaking down the silos that typically limit cross-functional work. It’s also worth noting that positive emotions increased 46 percent for individuals using AI and 64 percent for AI-augmented teams, while anxiety and frustration dropped roughly 23 percent (Candelon, 2025).

Innovate is at the frontier of AI implementation. This approach involves a complete reimagining of jobs, workflows, and even organizations, using AI to create value in ways that weren’t previously possible.

Isomorphic Labs, a company spun out of Google’s DeepMind, is transforming pharmaceutical R&D through computational AI. Instead of the traditional trial-and-error approach to testing new compounds in the lab, AI allows them to model molecule interactions digitally at scale (Callaway, 2024). Only the ones that are validated through digital simulation move on to lab tests. This process enables exponentially more testing, reduces development time, and allows exploration of drugs for rare diseases that wouldn’t have been financially viable under the traditional process.

Wharton professor Ethan Mollick, one of the most prominent researchers studying AI’s impact on work, has argued that the displacement of human workers by AI is not an inevitable consequence of the technology. Instead, he says, it is a leadership choice. Organizations that approach AI with a narrow lens will default to the efficiency trap. Those with a broader vision will recognize this moment for what it is—a rare opportunity to fundamentally redesign work around what allows people to thrive (Mollick, 2024).

This is not simply a question of technology. It is a question about what kind of organizations we want to build and what kind of work we believe people deserve. Will we use this extraordinary technology to amplify what makes us human, or will we allow it to quietly erode the very things that make work worth doing?

The choice belongs to us.

ActivTrak. (2026). 2026 state of the workplace: AI adoption and workforce performance benchmarks. ActivTrak Productivity Lab. https://www.activtrak.com/resources/state-of-the-workplace/

Adams, W., & Myles, T. (2025). Meaningful work: How to ignite passion and performance in every employee. PublicAffairs.

Beam AI. (2025). Why 42% of AI projects show zero ROI. Beam AI Agentic Insights. https://beam.ai/agentic-insights/why-42-of-ai-projects-show-zero-roi-(and-how-to-be-in-the-58-)

BetterUp Labs. (2025). Pilots and passengers: The mindset divide shaping AI adoption. BetterUp. https://grow.betterup.com/pilots-passengers

Brynjolfsson, E., Li, D., & Raymond, L. R. (2023). Generative AI at work (NBER Working Paper No. 31161). National Bureau of Economic Research. https://www.nber.org/papers/w31161

Callaway, E. (2024). Major AlphaFold upgrade offers boost for drug discovery. Nature, 629(8012), 509–510. https://doi.org/10.1038/d41586-024-01383-z

Candelon, F. (2025, October 31). The 'cybernetic teammate': How AI is rewriting the rules of business collaboration. Fortune.

Challapally, A., et al. (2025). The GenAI divide: State of AI in business 2025. MIT. https://www.artificialintelligence-news.com/wp-content/uploads/2025/08/…

Dell'Acqua, F., McFowland, E., III, Mollick, E. R., Lifshitz-Assaf, H., Kellogg, K., Rajendran, S., Krayer, L., Candelon, F., & Lakhani, K. R. (2023). Navigating the jagged technological frontier: Field experimental evidence of the effects of AI on knowledge worker productivity and quality. Harvard Business School Technology & Operations Management Unit Working Paper, (24-013).

Haun, L. (2025, May 19). Klarna claimed AI was doing the work of 700 people. Now it's rehiring. Reworked. https://www.reworked.co/employee-experience/klarna-claimed-ai-was-doing…

Kosmyna, N., Hauptmann, E., Yuan, Y. T., Situ, J., Liao, X.-H., Beresnitzky, A. V., Braunstein, I., & Maes, P. (2025). Your brain on ChatGPT: Accumulation of cognitive debt when using an AI assistant for essay writing task. arXiv. https://arxiv.org/abs/2506.08872

Liu, Y., Wu, S., Ruan, M., Chen, S., & Xie, X.-Y. (2025). Research: Gen AI makes people more productive—and less motivated. Harvard Business Review. https://hbr.org/2025/05/research-gen-ai-makes-people-more-productive-an…

McNamara, S., & Thornton, J. (2025). Artificial intelligence replacement dysfunction (AIRD): A call to action for mental health professionals in an era of workforce displacement. Cureus. https://doi.org/10.7759/cureus.93026

Mollick, E. (2024). Co-intelligence: Living and working with AI. Portfolio/Penguin.

Moreno, A. (2025, July). How AI is redefining managerial roles. Harvard Business Review. https://hbr.org/2025/07/how-ai-is-redefining-managerial-roles

Pertama Partners. (2026). AI project failure statistics 2026: The complete picture. https://www.pertamapartners.com/insights/ai-project-failure-statistics-…

Shah, D. (2024). Keynote address. INBOUND 2024, Boston, MA. https://www.youtube.com/watch?v=pPQngmSEIe0

Stanford Institute for Human-Centered Artificial Intelligence. (2025). What workers really want from artificial intelligence. Stanford University. https://hai.stanford.edu/news/what-workers-really-want-from-artificial-…

Sukharevsky, A., Kerr, D., Hjartar, K., Hämäläinen, L., Bout, S., & Di Leo, V. (2025). Seizing the agentic AI advantage. McKinsey & Company, QuantumBlack. https://www.mckinsey.com/capabilities/quantumblack/our-insights/seizing…

Sun, L., Hu, R., & Su, H. (2026). Unlocking human potential in the AI age: How employee-AI collaboration transforms work engagement through dual psychological pathways. Frontiers in Psychology, 16. https://doi.org/10.3389/fpsyg.2025.1705671

Xu, F., & Qin, J. (2026). Promoting or preventing? The influence of artificial intelligence awareness on employee job crafting. Journal of Industrial Relations. https://doi.org/10.1177/00218863251360852

There was a problem adding your email address. Please try again.

By submitting your information you agree to the Psychology Today Terms & Conditions and Privacy Policy

© Psychology Today