- Change, IT Basics
- Change Management, IT
- 13-minute read
Dr. Yvonne Avaro
AI has been around for a long time. But only 0.04% of the world’s population actually uses it. That’s roughly the same number of people as in Switzerland and Austria combined. What’s holding the rest back? And what does this mean for companies that are currently trying to get their workforce on board? Dr. Yvonne Avaro has combed through the research and identified six factors that determine whether AI adoption in companies succeeds or fails. From perceived benefits to trust to corporate culture: The AI Adoption Flywheel shows where organizations need to start.
💡New here? In the FAQ section at the end of the article, you'll find answers to common questions, including: What are the most common reasons why AI adoption fails in companies? How long does it take? And what specific steps can executives take?
Table of contents
1. Global AI Adoption: AI Is Here!
A look at the numbers shows just how far the world really is from true AI adoption. But before we get to that, let’s first establish a common understanding of the term “AI adoption.”
How do you define AI adoption?
AI adoption refers to the actual, regular use of AI tools. It’s not about trying them out occasionally, but about genuinely integrating them into everyday life: into workflows, decision-making processes, learning processes, or creative activities.
What Is the State of Global AI Adoption?
AI is here and readily available to most people. Yet many are still hesitant. Figure 1 illustrates just how hesitant they are. Figure 1 shows 2,500 points. Each point represents 3.2 million people. Together, that amounts to 8.1 billion people on Earth (as of February 2026).
What do the different colors mean?
- Gray: 6.8 billion people who have never used AI. 84% of the world's population.
- Green: 1.3 billion free chatbot users. People who open ChatGPT every now and then, ask a question, and move on.
- Yellow: 15 to 25 million people who pay for AI each month. About 0.3%.
- Red: A tiny strip. 2 to 5 million people who actively integrate AI into their workflows, build products using it, and work with it every day. That’s 0.04% of the world’s population.
What does the graph tell me?
Only 0.04% of the world’s population uses AI effectively to build things. At first glance, that might not sound like much. But that still amounts to as many as 17–30 million people worldwide, if you include paying users.
And at the same time: 6.8 billion people have never used AI. The potential for genuine, widespread AI adoption is enormous. And that’s exactly why it’s worth understanding what actually drives adoption.
2. How I Explored the Question of What Drives AI Adoption
What really drives AI adoption? That was the question I started my research with.
So I did something I hadn't done since my doctoral dissertation: a proper literature review. Journals, studies, research papers. The whole shebang.
The result is a model with six factors that determine whether people truly adopt AI in their day-to-day work or not. It is visualized as a wheel, with AI adoption at the center and the six factors as segments around it.
3. The AI Adoption Flywheel: The Six Factors for Successful AI Adoption
An extensive literature review I conducted shows that there is no single reason why people accept or reject AI. There are six factors that interact. Some are related to the individual, and some to the organization. All six can be influenced.
The Flywheel provides an overview of these factors (see Figure 2). At the center is AI adoption. Surrounding it are the six segments that determine whether adoption succeeds or fails. No single factor stands alone. And no company can afford to ignore any of them. The six factors are described in more detail below.
3.1. Perceived Value
In adoption research, perceived usefulness is considered the strongest single predictor of the actual use of new technologies. Davis’s (1989) Technology Acceptance Model (TAM) identifies “perceived usefulness” as a key determinant of intention to use: individuals adopt a technology when they are convinced that it measurably improves their work performance (Davis, 1989, https://doi.org/10.2307/249008).
This basic assumption has been confirmed in AI-specific research. In their comprehensive literature review on generative AI, Dwivedi et al. (2023) show that perceived benefits—particularly in the form of increased productivity, time savings, and improved work quality—are consistently identified as a dominant factor influencing adoption decisions. Without this subjective perception of benefit, even technically mature software remains unused (Dwivedi et al., 2023, https://doi.org/10.1016/j.ijinfomgt.2023.102642).
At the European level, Eurofound (2025) confirms this finding: Employees who perceive a concrete benefit for their own work show significantly higher adoption rates than those who view AI primarily as an abstract technological innovation. The report emphasizes that an individual’s perception of the benefits depends more on how closely a tool relates to their specific tasks than on general attitudes toward digital technologies (Eurofound, 2025, https://www.eurofound.europa.eu/en/publications/2025/impact-artificial-intelligence-work).
In summary, it can be said that without perceived personal value, no sustainable behavioral change will take place. No tool—regardless of its objective capabilities—can replace this step. The practical implication for organizations is therefore clear: AI implementations must be geared toward concrete, task-related value propositions from the very beginning.
3.2 Digital Skills & AI Literacy
The OECD (2024) defines AI literacy in this sense as the ability to understand AI systems, evaluate them critically, and integrate them effectively into one’s own work context. The report shows that employees with higher AI literacy not only demonstrate a greater willingness to adopt AI tools but also use them more effectively and derive measurably greater benefits from them (OECD, 2024, https://www.oecd.org/en/publications/2024/ai-skills-and-labour-market.html).
In its regular reports on the Digital Economy and Society Index (DESI), the European Commission reaches a similar conclusion: In countries where the population has higher basic digital literacy, the AI adoption rate—in both personal and professional contexts—is significantly above the EU average. At the same time, the reports identify a structural skills gap across occupational groups, age groups, and educational levels as a major barrier to widespread societal adoption of AI (European Commission, DESI 2024, https://digital-strategy.ec.europa.eu/en/policies/desi).
The psychological mechanism behind this connection is particularly relevant: A lack of understanding leads to feelings of loss of control and—as discussed in more detail in the following section—fear of negative consequences. AI literacy directly counteracts this mechanism: It lowers the perceived barrier to entry, strengthens the sense of self-efficacy, and creates the cognitive foundation needed to evaluate the perceived benefits (see Section 3.1) in the first place.
For organizations, this leads to a clear course of action: AI implementations without accompanying skills development will structurally fall short of their potential, regardless of how powerful the technology used is.
3.3. Openness to Innovation (Openness to New Ideas)
Individual openness to new technologies—referred to as “innovativeness” in adoption research—is one of the most robust behavioral predictors of early and sustained AI use. In his influential Diffusion of Innovations Theory (1962, revised 2003), Everett Rogers describes five categories of adopters, ranging from “Innovators” to “Laggards,” and shows that the willingness to try something new depends largely on how individuals assess uncertainty, complexity, and the perceived change brought about by an innovation (Rogers, 2003, https://doi.org/10.4324/9780203031650). “Innovators” and “Early Adopters” are not primarily characterized by superior expertise, but rather by a fundamental curiosity and a greater willingness to take risks when dealing with the unknown.
Recent research findings confirm that this mechanism is also central to AI adoption in the workplace. In an empirical study, Shao et al. (2025) examine the interplay between personality traits and AI adoption behavior and show that openness to experience—an established personality construct from the Five-Factor Model—correlates significantly with the willingness to actively integrate AI tools into everyday work. Notably, the study finds that formal training alone does little to increase this willingness: the decisive impetus for learning often arises from self-initiated experimentation, not from top-down mandated training (Shao et al., 2025, https://www.mdpi.com/2076-328X/15/4/491).
The latest SHRM report, “Navigating AI in the Workplace 2026,” also supports this finding at the organizational level: Employees who were allowed to test AI tools on their own initiative and in an exploratory manner showed a significantly higher frequency of use and a broader range of applications after six months than those who were introduced to the tools exclusively through structured training programs. The report explicitly recommends creating organizational spaces for experimentation where curiosity can be explored without performance pressure (SHRM, 2026, https://www.shrm.org/topics-tools/research/navigating-ai-in-the-workplace/full-report).
Openness to innovation is therefore not merely a dispositional trait that organizations must take into account when selecting personnel; it can also be shaped by culture. Those who establish a culture of learning that rewards curiosity and tolerates mistakes made while experimenting create the structural conditions necessary for hesitant observers to become active users.
3.4 Job Relevance (Relevance to One's Own Work)
In its comprehensive analysis of the impact of AI on the European labor market, Eurofound (2025) shows that adoption rates vary significantly by occupational group and job profile. Knowledge workers in fields such as communications, marketing, consulting, data analysis, and software development consistently report higher usage rates than workers in operational or manual roles. The decisive factor here is the structural compatibility between the capabilities of generative AI systems—text generation, pattern recognition, and information synthesis—and the dominant types of tasks associated with a professional role (Eurofound, 2025, https://www.eurofound.europa.eu/en/publications/2025/impact-artificial-intelligence-work).
The OECD (2024) further refines this finding: It is not just the occupational group, but the specific mix of tasks within a role that determines the likelihood of adoption. Employees whose daily work involves a high proportion of cognitive, language-based, or analytical subtasks identify opportunities for the productive use of AI more quickly and more frequently. Conversely, willingness to adopt AI remains low when there is a perceived mismatch between the tool’s capabilities and the reality of the work, even when there is a general openness to the technology and digital competence (OECD, 2024, https://www.oecd.org/en/publications/2024/ai-skills-and-labour-market.html).
This finding has direct practical implications for how AI is implemented in organizations. An undifferentiated rollout strategy that targets all employees with the same tools and the same promises of benefit fails to account for the reality of heterogeneous job profiles. More effective are role-specific onboarding scenarios that demonstrate exactly where AI makes a direct difference in employees’ daily work—not as an abstract promise for the future, but as a solution to a familiar, everyday task.
The key question regarding AI adoption is therefore not “Is AI good?”—that’s a societal debate. The relevant question at the individual level is: “Does this tool help me with what I need to do tomorrow morning?”
3.5. Risk Perception & Trust (Trust and Risk Perception)
Trust in AI systems and the perception of associated risks act as active barriers to adoption, even when willingness to use them, competence, and task relevance are generally present.
The Deloitte European Trust in AI Studies (2024/2025) identify a consistent set of concerns: data protection, data security, the reliability of AI outputs, algorithmic biases, and individual concerns about negative consequences for one’s own employment. It is noteworthy that these concerns do not decrease linearly with technical knowledge—even well-informed professionals express substantial reservations when organizational frameworks are unclear (Deloitte, 2024/2025, https://www.deloitte.com/de/de/pages/risk/articles/european-ai-trust-study.html).
The OECD shows that trust depends largely on two factors: the transparency of the AI system itself and institutional trust in the organization’s governance framework. Companies with clearly communicated AI policies that are demonstrably followed report significantly higher adoption rates (OECD AI Policy Observatory, https://oecd.ai/en/policy-areas/trust-in-ai).
The key finding: Trust is not built through communication alone. Concrete measures—such as transparent rules of use, clearly defined responsibilities, and leadership that visibly puts these practices into action—are more effective than any mere commitment to “safe AI.” Governance is therefore not a bureaucratic byproduct of AI implementation, but rather a direct measure to drive adoption.
3.6. Organizational Support & Culture (Organizational Support)
Individual willingness alone is not enough. Even self-motivated, competent employees tend to use AI less frequently and more superficially if the organizational environment does not take a clear stance on the matter.
In their State of AI Reports (2024/2025), McKinsey & Company show that companies with active leadership support, structured training programs, and explicit AI guidelines achieve significantly higher adoption rates than those that implement AI informally and without a framework. The signal effect of leadership is particularly relevant here: When managers use AI tools visibly and thoughtfully, the team’s willingness to use them demonstrably increases (McKinsey, 2024/2025, https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai).
The OECD expands on this finding from a cultural perspective: Psychological safety—that is, the shared belief that trying new things and making mistakes will not result in negative consequences—is an independent predictor of AI adoption in the workplace. Employees who know that experimentation is explicitly encouraged are able to establish productive usage routines more quickly and tap into broader areas of application (OECD Workplace AI Studies, https://oecd.ai/en/policy-areas/work).
Organizational support thus works on two levels simultaneously: It lowers the perceived barrier to entry and increases the willingness not only to try out AI on an ad hoc basis, but also to integrate it permanently into workflows.
Read more in this wiki about IT change management and how to make it a success.
4. Why should I look into the AI Adoption Flywheel to promote AI adoption in my company?
No matter how much a company invests in AI tools, if its culture punishes mistakes, this will have a direct impact on adoption. Similarly, even the most open corporate culture is of little use if employees don’t see any personal benefit in the tools.
The six influencing factors from sections 3.1 through 3.6 together represent the dimensions of the AI Adoption Flywheel. Six questions—one for each dimension—show where organizations stand today and where there is a specific need for action:
- Perceived Value: Do our employees see real value in their day-to-day work?
- AI Literacy: Do you have the basic digital skills needed to use AI effectively?
- Openness to Innovation: Is there room to experiment without fear of making mistakes?
- Job Relevance: Is AI even relevant to these roles?
- Trust & Risk Perception: Do people trust the systems and the framework we have established?
- Organizational Support: Does leadership actively and visibly support the implementation?
AI adoption isn't a technical problem. It's a human one. And it requires solutions on all six levels at the same time.
What does this look like in practice? In the following section, you’ll find a concise practical guide—a sort of etiquette guide for AI adoption in everyday business life.
5. AI Adoption Guide
What leaders and teams can do specifically, one dimension of the flywheel at a time.
1. Perceived Value: Making Value Visible
Concrete examples show how AI is already simplifying work within the company—not in abstract terms, but in a role-specific and practical way. An example from your own team is more effective than any glossy presentation. Even before the project begins, make a point of collecting internal success stories and actively sharing them. Define two to three specific use cases for each role before introducing tools. KPIs, KPIs, KPIs: Measure time savings and quality improvements wherever possible. Numbers build conviction.
2. AI Literacy: Building Competence Without Overwhelming Users
No one needs to know everything. It’s enough to understand just enough to get started. Start with short, hands-on sessions rather than multi-day training courses. Rely on peer learning: Those who already use AI can explain it to their colleagues better than any external trainer. Also, provide simple resources such as an internal wiki, short explanatory videos, and an FAQ page.
3. Openness to Innovation: Creating Space for Experimentation
Curiosity requires a sense of security. People who are afraid of making mistakes won’t try anything new. Make it explicitly clear that experimentation is encouraged—and feel free to repeat this message more than once. It also helps to set up safe testing environments where employees can experiment without fear of repercussions. And yes: Celebrate victories! Learning moments are successes.
4. Job Relevance: Introduce It in a Way That Fits Perfectly
6. Organizational Support: Leadership by Example
Employees observe what leaders do, not just what they say. Where AI tools are used in a visible and thoughtful manner at the leadership level, team engagement has been shown to increase. Making AI adoption a strategic topic in leadership discussions and goal-setting sends a clear signal. The same goes for allocating time and budget for professional development: Failing to provide both signals that it isn’t really important.
AI adoption isn't achieved through a single project. It results from many small, consistent decisions across all six levels, every day.
5. Conclusion & What Really Matters Now
AI, as a technological transformation, is driving a fundamental transformation of the workforce and workflows.
Success doesn't come from algorithms or models. It arises when people are empowered to harness the potential of AI in their daily work. That is the key difference between companies that truly implement AI and those that get stuck at the pilot project stage.
Underestimating training is the most costly mistake, but curiosity comes first.
This isn’t a contradiction: Many people discover AI through trial and error, and that’s a good thing. Curiosity is the gateway. But for that initial experiment to lead to sustainable adoption, more is needed: structured upskilling that fosters an understanding of processes, builds change management skills, and empowers people to deal with uncertainty. Those who invest only in tools but not in people will not see sustainable adoption.
Just get started instead of trying to plan everything perfectly.
Not every use case has to have a major business impact. That’s not the point. Use cases with low barriers to entry are a good place to start: vacation planning, meeting summaries, organizing notes from a half-day workshop. What matters is the sense of accomplishment. Once you’ve experienced what AI can do, you’ll stick with it.
AI adoption doesn't happen on its own.
The good news: The six factors in the AI Adoption Flywheel can be influenced. Perceived value can be communicated. AI literacy can be developed. Trust can be earned. Culture can be shaped.
Those who understand these factors and adapt them to their own teams and circumstances can give AI adoption a real boost.
Value doesn't come from algorithms or models. It arises when people are empowered to harness the potential of AI in their daily work. That is the key difference between companies that truly implement AI and those that get stuck at the pilot project stage.
For specific support with IT change management and adoption KPIs:
Schedule a discussion with Robert on this topic here. The conversation will cover IT change management—from stakeholder alignment and communication to adoption KPIs and sustainable integration into everyday work.
List of Sources and Further Reading
- Davis, F.D. & Granić, A. (2024): The Technology Acceptance Model: 30 Years of TAM, Springer Nature (A foundational work on the Technology Acceptance Model / TAM)
- Dwivedi et al. (2023): Generative AI Adoption Studies, available via ResearchGate and academic journals
- Eurofound (2025): Collective Bargaining on Artificial Intelligence at Work 🔗 https://www.eurofound.europa.eu
- OECD (2024): AI and the Future of Work https://www.oecd.org/en/publications/oecd-employment-outlook-2023_08785bba-en.html
- McKinsey & Company (2025): The State of AI: Agents, Innovation, and Transformation 🔗 https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai
- McKinsey & Company (2025): AI in the Workplace 🔗 https://www.mckinsey.com/capabilities/tech-and-ai/our-insights/superagency-in-the-workplace-empowering-people-to-unlock-ais-full-potential-at-work
- Deloitte (2024/2025): European Trust in AI Studies 🔗 https://www.deloitte.com/insights
- Rogers, E.M. (1962/2003): Diffusion of Innovations, 5th ed., Free Press, New York 🔗 https://www.researchgate.net/publication/257624104
- European Commission (2024): State of the Digital Decade 2024 / Digital Economy and Society Index (DESI) 🔗 https://digital-strategy.ec.europa.eu/en/policies/2024-state-digital-decade-package
- Shao et al. (2025): The Impact of AI Usage on Innovation Behavior at Work: The Moderating Role of Openness and Job Complexity, Behavioral Sciences 🔗 https://www.mdpi.com/2076-328X/15/4/491
- SHRM (2026): Navigating AI in the Workplace 2026 🔗 https://www.shrm.org/topics-tools/research/navigating-ai-in-the-workplace/full-report
- PwC / WEF (2025): Global Workforce Hopes and Fears Survey 2025 🔗 https://www.pwc.com/gx/en/issues/workforce/hopes-and-fears.html
_________________________________________________________________
This wiki is based on Yvonne Avaro's presentation at the Analytics Summer Apéro 2026. LINK: https://s-peers.com/analytics-analytics-summer-apero-2026-aufzeichnungen/
6. FAQs
The most common causes are a perceived lack of benefit, a lack of trust in AI systems, and a corporate culture that does not allow for experimentation. Many implementation projects fail not because of the technology itself, but because employees do not see any personal value in the tools or are afraid of the consequences of making mistakes.
According to Deloitte, most organizations need 12 months or more to resolve governance, training, trust, and data issues to the extent that AI usage can be scaled. The first measurable effects on productivity and work quality often become apparent after just a few weeks for individual teams, but widespread adoption is a longer process.
AI adoption refers to whether, how often, and how effectively employees use AI tools in their day-to-day work. It’s not just about the frequency of use, but also about whether AI is actually changing the way work is done, whether it influences decisions, accelerates processes, and creates real added value. AI integration goes a step further: it describes whether AI is systematically embedded in processes, workflows, and decision-making structures. Many companies have high adoption rates at the individual level but have not yet achieved true integration at the organizational level.
Knowledge workers in communications, marketing, consulting, analytics, and software development show the highest adoption rates. This is because AI tools directly support core tasks in these fields, such as content creation, data analysis, and code generation. The higher the job relevance, the faster the adoption.
AI literacy refers to an understanding of what AI systems are capable of, where their limitations lie, and how to use them effectively. People with AI literacy experiment more often, use tools more effectively, and overcome initial uncertainty more quickly. AI literacy is therefore one of the most important prerequisites for sustainable adoption.
Trust operates on two levels: trust in the technology itself (accuracy, data protection, transparency) and trust in the organization (clear rules, fair treatment, no fear of job insecurity). If either of these is lacking, people will be reluctant to use the tools, even if they are technically mature.
Leaders have a direct influence on five of the six factors in the AI Adoption Flywheel. Specifically, they can help by: visibly using AI themselves and talking about it, establishing clear guidelines and a governance framework, explicitly allowing experimentation, facilitating professional development, and celebrating quick wins within the team. Adoption follows leadership behavior.
AI adoption is change management. The factors in the flywheel reflect exactly what makes change processes succeed or fail: value, trust, security, and engagement.
