A Different Way of Thinking About Technological Change
Image: Depositphotos
There’s increasing excitement about Gen AI, especially GPT-5. Certainly, it appears that ChatGPT-5 improves on ChatGPT-4. It boasts stronger reasoning, a larger memory span for handling long or complex conversations, and smoother multimodal capabilities that integrate text and images more effectively. It may generate responses faster and with greater factual accuracy, reducing errors and hallucinations. GPT-5 also follows instructions more reliably, adapts tone and style more naturally, and produces concise, human-like answers. Overall, it appears to be a more accurate, efficient, and versatile model than GPT-4.
At the same time, new research reveals a widening gap between hype and real enterprise value. A recent article reviews MIT’s The GenAI Divide: State of AI in Business 2025, revealing that 95% of generative-AI pilot programs fail to deliver meaningful business outcomes, such as revenue growth or productivity improvements. Based on 300 real-world deployments, 150 executive interviews, and 350 employee surveys, the study highlights a pervasive “learning gap”—organizations struggle to integrate AI into workflows and set aligned goals.
The MIT report cites the following reasons for failure:
Shallow Pilots, Not Scalable Solutions
Cultural and Organizational Resistance
Poor Alignment with Business Goals
Weak Data and Infrastructure Foundations
Underinvestment in People and Processes
Projects focused on back-office automation and those leveraging external vendor tools fared much better than internally developed or marketing-oriented initiatives. The lack of value creation is not due to the quality of the AI models, but related to cultural issues, poor resource allocation, and inadequate change management. In other words, generative AI adoption falters less because of the technology itself and more because organizations lack the cultural readiness, disciplined resource allocation, and robust change management required for meaningful transformation.
While the methodology used in this study has drawn some criticism, the findings are directionally correct as there is a large gap between hype and reality when it comes to Gen AI.
Part of the problem is that many organizations continue to treat AI projects as primarily technical challenge. But it goes beyond that to leadership mindset and deeply ingrained habits. In many enterprises, leaders persist in viewing performance in the context of the organization chart - and that is a serious problem.
When leaders view performance mainly through the lens of the organization chart hierarchical positions and formal reporting lines are top of mind, rather than actual value creation, collaboration, or customer experience. This mindset reinforces silos, as leaders focus on how “their” unit appears rather than how work flows across functions. It drives a mentality where small AI pilot projects are defined inside departmental boundaries. It also risks overemphasizing positional authority instead of skills, contribution, and adaptability, which are critical in today’s rapidly changing, technology-driven environment. By tying performance to the org chart, leaders can unintentionally discourage cross-functional problem-solving, stifle innovation, and overlook the big picture view of deploying AI. In a nutshell, this view impedes deploying AI to transform the end-to-end processes and customer experiences that drive results.
The track record of AI in new product development illustrates the issue with management mindset. While companies have deployed AI effectively for incremental improvement in new product development, large scale change remains elusive. New product development is inherently cross-functional, involving marketing, R&D, operations, and other departments. Yet, AI solutions are often siloed within one function, limiting enterprise-wide impact. Data challenges compound the issue, as insights, design files, and production information are fragmented across legacy systems.
A new way of thinking about technological change is needed and it’s not about adding new roles to the C-suite - as that just reinforces an outdated mental model.
Effectively deploying AI requires fundamental change in several ways:
· AI implementation requires leaders to work across traditional silos with an emphasis on end-to-end process performance, instead of focusing on small projects inside departmental boundaries
· Greater management attention has to be devoted to creating customer value versus just cost reduction
· Careful consideration of ethical considerations and governance is needed
· Going beyond the hype and developing a more profound understand AI capabilities and human skills and psychology
The following are some other actions to go beyond the hype of AI:
· Focus on projects that directly support revenue growth, or customer value, rather than chasing hype or “nice-to-have” pilots. Clearly define success metrics.
· Encourage building customer journey maps and high-level process diagrams. Then, provide training, communicate transparently about AI’s role, and help to involve employees early. Building trust and capability reduces resistance and accelerates adoption.
· Ensure high-quality, accessible data and redesign workflows to integrate AI smoothly. Technology succeeds only when embedded in processes that create measurable outcomes.
· Launch targeted pilots in high-impact areas, measure results rigorously, and scale proven use cases across the enterprise.
· Break down silos by bringing together IT, data science, operations, and business units. AI success depends on collaboration across technical and domain experts.
· Set up clear guardrails around data privacy, bias, and accountability.
· Treat AI as a learning journey, not a one-time implementation. Regularly assess outcomes, refine models, and adapt strategies as business conditions and technologies evolve.
The combination of the above mentioned shift in thinking and proposed actions can drive greater success with AI.
At the Snowflake Summit in San Francisco on June 2, 2025, OpenAI CEO Sam Altman compared GPT-5 to having a “team of Ph.D. level experts in your pocket.” This might be a slight exaggeration in light of current performance. Thinking differently about technology is the best way to decrease the size of the gap between hype and AI’s actual performance.
Andrew Spanyi is the editor at Cognitive World and the founder of Spanyi International Inc. He is the author of three books on process management and operational leadership. He has written over a hundred of articles and is a BPM Fellow with the Association of Business Process Management Professionals. He was formerly an adjunct professor at Babson College. His current research interest is on redesigning the oncology clinical trial process. Andrew is a member of the Cognitive World Think Tank.
LinkedIn