Why Some AI Pilots Win, and Others Stall

Many companies have joined the AI race. Over the past two years, we’ve seen an explosion of pilots: generative models for marketing copy, chatbots for customer service, copilots for developers. According to an MIT report, despite the momentum, only about 5 percent of organizations have successfully scaled AI pilots into measurable business impact. 

What separates the few that scale from the many that stall? 

An article in Harvard Business Review sheds light on this question. Their central argument is that successful AI adoption hinges less on algorithms and more on leadership. Companies that scale AI pilots do so because they have leaders who can connect technology with strategy, embed it into workflows, and build the trust needed for adoption. They call these leaders AI shapers. 

Why Technical Expertise Alone Is Not Enough 

Think of Johnson & Johnson, one of the largest healthcare companies in the world. Over the course of a few years, it ran nearly 900 AI pilots. By sheer numbers, it should have been a success story. But most of those pilots went nowhere. Instead of declaring victory on volume, J&J made a difficult but crucial pivot: it shut down the majority of pilots, moved governance closer to business units, and doubled down on the handful that actually created measurable value. 

This example illustrates the heart of the issue. Talent and experimentation are necessary but not sufficient. What matters is the ability to sift through experiments, choose what aligns with business goals, and scale only the use cases that matter. That is a leadership challenge, not a coding challenge. 

The Five Behaviors of AI Shapers 

The findings show five leadership behaviors that consistently show up in companies that succeed with AI pilots. They call this the SHAPE index: strategic agility, human centricity, applied curiosity, performance drive, and ethical stewardship.

Image 1: The SHAPE Index: Five Behaviors of AI Shapers (Havard Business Review) 

S-Strategic Agility 

Leaders with strategic agility treat pilots as experiments rather than rigid roadmaps. They ask tough questions early: Does this create business value? Do we have criteria for when to pivot? 

Consider a global logistics company that piloted a generative AI tool for delivery route optimization. Early results fell short of expectations. Instead of increasing investment into the tool, leadership redirected resources into predictive maintenance AI, which quickly showed cost savings across the fleet. The decision wasn’t about abandoning AI, but about adjusting the course to capture value. 

Organizations that lack this agility often defend projects because they’ve already invested heavily in them. That mindset leads to inefficient use of resources and lost opportunities for value creation.  

H-Human Centricity 

Trust sets the pace of AI adoption. When employees feel sidelined or threatened, pilots lose momentum. Leaders with human centricity ask a different question: How can AI enhance people’s contributions? 

At a European bank, leadership introduced a generative AI compliance assistant. Rather than imposing it top-down, they invited compliance officers into the design process. Employees came to see the tool as a support for their work, not a replacement, and adoption followed. By contrast, firms that announce “AI will make us more efficient" without addressing employee concerns often encounter resistance that slows progress. 

This reflects a broader workplace dynamic. Stanford psychologist Jamil Zaki has described what he calls an “empathy crisis” in today’s organizations. AI can either intensify that challenge or become a tool for empowerment, depending on how leaders choose to introduce it. 

A-Applied Curiosity 

Generative AI is evolving rapidly, with fresh applications appearing almost daily. The leaders who succeed are those who combine curiosity with discipline. They launch small, low-cost pilots with clear learning objectives, then double down on the few that deliver tangible value. 

Applied curiosity means exploring widely but scaling selectively. It’s about testing many ideas, capturing lessons quickly, and channeling resources into the ones that create measurable business impact. 

Without this discipline, organizations risk spreading themselves too thin, running pilots that never connect back to real priorities. 

P-Performance Drive 

Many organizations equate activity with progress, emphasizing the number of pilots launched rather than the value created. Leaders with true performance drive shift the focus toward outcomes, whether in revenue growth, cost efficiencies, or customer experience, rather than vanity metrics. 

Johnson & Johnson offers a useful example. By streamlining its portfolio of pilots and scaling only those that demonstrated clear ROI, the company avoided unnecessary experimentation and concentrated resources on measurable impact. 

E-Ethical Stewardship 

Generative AI introduces new risks, including hallucinations, bias, and potential intellectual property exposure. Ethical stewardship means addressing these challenges from the very beginning by embedding governance, transparency, and oversight into the workflow.

For example, one media company built “red team” testing into every generative AI deployment, auditing outputs for accuracy and bias before going live. This not only reduced reputational risk but also strengthened trust with employees and stakeholders, creating the foundation for faster scaling.

By contrast, organizations that postpone governance until after issues emerge often find themselves struggling to catch up.

So What Leaders Can Do Differently? 

Moving from pilots to scale requires deliberate choices. Four stand out: 

• Assess where your leadership stands against the SHAPE framework. Without diagnosis, organizations misplace talent and misjudge readiness. 

• Hire for the capabilities that are hardest to teach, particularly strategic agility and applied curiosity. Technical skills are abundant; shaping skills are scarce. 

• Develop leaders who already show shaper behaviors by giving them visibility, stretch assignments, and targeted support. Move beyond generic bootcamps and focus on the few leadership skills that matter most. 

• Role model adoption at the top. Employees will not embrace AI if their leaders don’t use it themselves. Leaders who integrate AI into their daily decisions send a powerful signal that adoption is a shared priority, not a delegated task. 

  
  

Conclusion 

AI success will not be won in research labs alone. It will be won in workflows, in strategy rooms, and in the daily choices leaders make to guide adoption across the enterprise. The question is not how many pilots you launch, but whether you have the leadership to scale the ones that matter. 

The companies that cultivate AI shapers won’t just join today’s 5 percent of winners—they will define the playbook for successful AI adoption in the years ahead. The choice is yours: will your organization shape AI, or be shaped by it? 

References: 

Estrada, S. (2025, August 18). MIT report: 95% of generative AI pilots at companies are failing. Yahoo Finance. https://finance.yahoo.com/news/mit-report-95-generative-ai-105412686.html?_guc_consent_skip=1757847197 

Van Den Broek, R., Hellauer, S., & Wang, D. (2025, September 12). What Companies with Successful AI Pilots Do Differently. Harvard Business Review. https://hbr.org/2025/09/what-companies-with-successful-ai-pilots-do-differently?ab=HP-hero-latest-2  

. 

👉Interested in driving innovation with VinVentures? Share your venture with us HERE.