Part 2 of 5 in the "Why AI Projects Fail" Series

Image 1. A focused team analyzing data and documents in an office environment, with AI-driven insights on the screen. Credit - Vertex.

In Part 1 of this series, we explored why 80% of AI projects fail, not because of the technology itself, but because organizations are Complex Adaptive Systems (CAS)—interconnected, evolving environments where AI must be carefully integrated to thrive. People are complex.

“AI Success Hinges on People, Not Just Technology”

Now, let’s focus on the key factor in AI success: your employees. Your most experienced and insightful team members may initially resist AI adoption — not because they are being difficult, but because they spot risks and inefficiencies that leadership might overlook. The good news? Their insights are the key to making AI work better, faster, and more effectively.

AI solutions are marketed as tools to improve productivity, enhance decision-making, and streamline operations. Yet, adoption often stalls when AI tools are introduced in real workplaces. Employees may hesitate to use them, override automated suggestions, or struggle to integrate AI into their daily workflows.

A study by Lucid Software found that 33% of workers consider resistance to change a major challenge when implementing AI, as it disrupts established workflows. (Lucid Software)
According to research by Slack, AI adoption rates among U.S. workers have slowed, with only a 1% increase over three months, suggesting hesitation in embracing AI technologies since the burden to learn AI is put on workers. (HR Dive)
The Federal Reserve reports that 20% to 40% of workers use AI in the workplace, with higher rates in specific occupations like computer programming. (Federal Reserve)

Let’s explore how organizations can turn resistance into collaboration in more detail than we had in Part 1.

1. The Role of Expertise and Control in AI Adoption

Experienced employees have built their expertise over years—sometimes decades. When AI changes established workflows, they may worry about losing control over processes where their judgment is essential.

Example: A Fortune 500 finance department implements an AI-driven expense approval system. Senior accountants, who previously made nuanced judgment calls, now rely on AI flags without clear explanations. Their expertise is still needed, but AI doesn’t account for every real-world complexity, and so the accountants are wary. Additionally, AI makes learning for junior accountants hard because they are not exposed to non-standard situations.

Instead of viewing experts as obstacles, organizations should involve them early in designing AI solutions. This allows experts to shape how AI integrates with existing systems and refine AI models through their knowledge and experience. Make sure at least one, and up to three experts are included in the initial AI solution design.

2. Employees Want AI to Be Trustworthy

Employees understand the context, exceptions, and edge cases that AI may initially overlook.

“When AI lacks transparency, it’s natural for teams to question its reliability.”

Including employee feedback early in AI system design is crucial for identifying edge cases and potential issues that could erode trust. When employees spot scenarios where AI might fail or make questionable decisions, organizations can address these concerns proactively rather than discovering them after deployment. By incorporating this feedback during development, organizations create more reliable and trustworthy AI tools that employees are more likely to embrace and use effectively.

Example: Amazon’s AI-powered hiring tool was trained on past hiring data, leading it to downgrade resumes from women. Employees recognized the bias before leadership. Their skepticism wasn’t resistance, it was insight. Gone uncaught, this bias may have led to expensive consequences. A study by Reuters (2018) revealed that Amazon discontinued the tool after discovering the bias, highlighting the need for continuous monitoring and human oversight in AI-driven decision-making (Reuters).

By incorporating employee feedback, AI can become more trustworthy and fair.

3. AI’s Fit into Real-World Workflows

If AI tools don’t match how employees actually work, those same employees will find ways to bypass the AI.

“Employees are not resisting innovation — they are adapting to maintain efficiency within their familiar workflows.”

The solution is not to enforce AI, but to design AI integrations alongside the people who will be using it, ensuring that it enhances their capabilities rather than disrupting established processes.

Example: A global sales team adopts an AI-powered CRM with automated deal scoring. Sales reps find the scores inconsistent and continue tracking deals manually in spreadsheets. Instead of forcing AI adoption, leadership can refine the model by integrating sales team feedback, ensuring AI improves their sales over time.

By tracking how AI solutions are used by people, continuous improvement can lead to more effective and widely-adopted tools that truly enhance employee capabilities while respecting established workflows.

The Science Behind Human-AI Collaboration

Resistance to AI is often rational and not reflexive. Employees spot potential flaws and unintended consequences that leaders miss. AI disrupts workflows and raises concerns about bias and job security. By addressing these valid concerns during implementation, organizations can transform hesitation into collaboration and successful adoption.

Here’s why employees may hesitate, and how organizations can turn hesitation into buy-in:

Cognitive Load and Change Management

Cognitive Load is the mental effort required to process information. When learning new AI tools, employees can become overwhelmed, leading to fatigue and resistance. If AI demands extensive retraining or disrupts workflows without clear benefits, employees often revert to familiar methods. Research shows that excessive cognitive load impairs learning and decision-making. Organizations should implement AI gradually with clear training to help employees adapt at a manageable pace. (Sweller, 1988)

Image 2: A professional working a late at night, reviewing documents under a desk lamp, showing the cognitive load and mental effort required when adapting to new AI systems in the workplace. Credit - Vertex

Psychological Ownership Drives Adoption

Psychological Ownership is when employees feel personally invested in their work and tools. This sense of ownership increases engagement, motivation, and commitment. Research shows it improves performance, job satisfaction, and innovation. However, AI can disrupt this if implemented without employee input. Studies by the Academy of Management (Pierce et al., 2001) and others (Lu & Zhao, 2017) show that involving employees in AI implementation—through feedback, customization, and expertise augmentation—leads to better adoption.

AI Can Benefit from Employee Insight

AI models are only as good as the data they are trained on. Employees often spot biases and blind spots first, as they work directly with real-world processes that AI may misinterpret. Their insights provide crucial context that helps refine AI outputs, ensuring the technology aligns with business needs and avoids unintended consequences.

Research shows that employee feedback improves AI systems. AI needs ongoing human input to work well (Ghosh, et al, 2018) and human insights build trust (Ransbotham, 2020). Companies that include employee feedback in AI development create better systems that foster trust in AI-driven decision-making.

How to Make AI a Trusted Partner

1. Involve Employees in AI Development

A 2024 McKinsey survey (McKinsey, 2024) found 91% of respondents felt unprepared for responsible AI implementation. Success in AI implementation requires inclusivity through structured end-user involvement. An excellent framework to include end-user involvement is through Design Thinking—where employees co-create AI solutions through ideation, prototyping, and feedback. This approach ensures AI complements workflows and builds trust. Forrester research (Forrester, 2018) showed Design Thinking leads to 50% faster adoption and 30% higher satisfaction when designing systems. By involving employees, businesses significantly improve AI implementation success.

2. Successful AI Complements Human Expertise

AI and humans excel at different things, making them ideal partners when properly integrated.

AI is particularly good at:

processing large amounts of data quickly
automating repetitive tasks
identifying patterns
making predictions based on historical data.

These strengths make AI an excellent tool for efficiency-driven processes such as fraud detection, customer service automation, and supply chain optimization.

On the other hand, humans are good at:

creativity
intuition
emotional intelligence
ethical reasoning

These qualities remain challenging for AI to replicate. Humans excel in decision-making that requires judgment, contextual understanding, and ethical considerations.

Image 3. Choices in leveraging strengths in AI as well as human capabilities. Credit: Napkin

For example, in healthcare, AI helps radiologists by flagging potential abnormalities, but final diagnoses remain with doctors. This collaborative model leads to better adoption and trust, particularly by the patients the doctors serve.

“Design AI to perform repetitive and high-volume tasks, while leaving strategic decision-making in the hands employees.”

In another example, in the legal field, AI-powered tools assist lawyers by scanning and analyzing large volumes of documents for relevant case law, but final legal arguments and strategy decisions remain with human attorneys.

This combination enhances efficiency while preserving human judgment, ensuring AI augments rather than replaces expert knowledge and experience. For AI to succeed in the workplace, it needs to enhance human strengths rather than replace them, ensuring that employees are empowered rather than sidelined.

3. Ensure Transparent AI Rollouts

Lack of clarity in AI implementation creates uncertainty about employee roles and job security. Without proper guidelines, employees may disengage from using AI tools effectively. Research shows that ambiguous expectations and change management reduce workplace productivity. A study in the International Journal of Organizational Analysis (Silva et al., 2024) confirms that role ambiguity hurts employee engagement and satisfaction, leading to lower performance.

Organizations implementing AI need to provide detailed explanations, structured training programs, and ongoing communication to build trust and encourage adoption. AI implementation needs to clearly define:

What AI will and won’t do.
How AI decisions are made.
How employees can provide feedback to refine AI models.

Conclusion: AI Success Depends on People

Your best employees are not obstacles to AI adoption—they are your greatest resource. When organizations collaborate with employees in the design of complex systems, AI becomes an asset that enhances expertise, not a tool that disrupts it. However, making this transition requires a structured approach, deep expertise, and a clear strategy.

Fyve Labs specializes in helping businesses bridge the gap between AI and human workflows, ensuring AI tools are effectively integrated into complex organizational structures. By leveraging methodologies like:

Private Company AI Chat
Design Thinking AI Product Workshops
Human-centered AI Custom Development

Fyve Labs helps companies transform AI resistance into AI adoption and AI optimization. Whether it’s starting you on your AI journey, refining AI models based on employee feedback, structuring training programs, or measuring AI impact with the right metrics, Fyve Labs provides the guidance necessary for AI success.

To learn how Fyve Labs can help your organization achieve seamless AI adoption, visit Fyve Labs.

Coming Up Next in this series:

Part 3: How telecom giants like AT&T, Verizon and T-Mobile cracked the AI code and what you can learn from them.

Need help making AI work with your people instead of against them? Contact us to learn how we can help.

The Human Factor: How Your Smartest Employees Can Help AI Succeed (and What Leaders Need to Know)