“Creating value for your company using Generative Artificial Intelligence (GenAI), like LLMs and chatbots, does not come from giving everyone access to these tools hoping that something magical will happen,” said Schmitt and Blohm.  In their research project, the two of them, along with Gregory Vial from HEC Montréal, interviewed 87 practitioners across 23 large organizations who have scaled GenAI usage successfully within their companies. They wanted to understand why the majority of companies were not seeing a strategic advantage in their businesses.

They found in most cases, giving employees access to general-purpose, large language models (LLMs) enhances their personal productivity, yet organisations as a whole are struggling see a strategic advantage or create value at scale when doing so. 

“The core problem is organizational, not technological” said the researchers. “Many firms have already given employees access to AI tools to improve personal productivity — for example, helping workers write emails, summarize reports, or generate ideas faster. However, these isolated productivity gains rarely create major competitive advantages. ”The team noted that real value comes when GenAI improves entire business processes across departments and functions, rather than helping individuals become more efficient. They identified three practices shared by companies that are successfully scaling GenAI.

The three difference maker for successful AI implementation:
1.    Broaden the scope of AI use. Instead of treating GenAI as a tool for one-off tasks, they integrate it into larger workflows and operational processes. This means thinking beyond “How can AI help this employee?” toward “How can AI improve how the organization operates overall?”

2.    Implement AI as an ongoing experiment rather than a finished product. GenAI systems require continual refinement, adaptation, and learning. These organizations succeed because they continuously adjust prompts, workflows, governance structures, and employee practices based on feedback and data.

3.    Abandon AI projects that do not create measurable value. Rather than continuing unsuccessful initiatives because of hype or sunk costs, these companies evaluate outcomes rigorously and redirect resources toward more promising use cases.

Schmitt and Blohm also note that traditional corporate structures often make these practices difficult. Many large companies operate in separate divisions with their own budgets, goals, and internal competition. Information and knowledge do not flow easily across departments, which limits the ability to scale successful AI solutions company-wide. Even if one department develops a useful GenAI application, other units may not adopt or even know about it.

The "AI Spine"

To solve this issue, the authors introduce the concept of an “AI spine.” This refers to a coordinating organizational structure that connects AI initiatives across departments and ensures knowledge sharing, governance, and strategic alignment. The AI spine acts as a bridge between technical teams and business units, helping organizations spread successful AI practices throughout the company instead of leaving them isolated within silos.

Another key insight is that the biggest challenge in generative AI adoption is not building the technology itself but redesigning the organization itself to support continuous experimentation, collaboration, and scaling. “Handling AI innovation sometimes is counter-intuitive because it is radically different to something that seems similar, such as IT innovation. IT innovation is outward looking and answers the question what product or service can we develop for a customer?” It identifies a challenge and try to solve the challenge with a specific goal in mind. AI innovation is more incremental and more inward looking and answers the question, “What can we improve?”

Companies that view GenAI as a company-wide transformation effort — rather than just another productivity tool — are more likely to achieve meaningful long-term returns on their investments.

Kevin Schmitt, Ivo Blohm and Gregory Vial’s more in-depth article on GenAI and their study Create Generative AI Value at Scale is available in MIT Sloan Management Review June 2026 edition.