What happened
A recent insight from a tech leader emphasizes that companies need to develop their own AI capabilities rather than relying solely on external models accessed via APIs. This shift in focus highlights the importance of the learning processes that surround the AI model, which contribute significantly to a company's intellectual property. Despite the buzz around implementing AI, many organizations face challenges that stem from neglecting the foundational elements necessary for successful integration.
Why it matters
Many mid-sized companies that invest in AI find that their excitement fades quickly after implementation. Despite having a budget for AI and selecting a model, they often experience stagnant or declining usage within months. The problem lies in the failure to build a supportive framework around the technology, leading to outputs that are rarely acted upon. A new perspective suggests that AI's value isn't additive but multiplicative, meaning that several components must work together for successful outcomes. If any of these components are missing, the overall effectiveness drops to zero.
Context
The concept of a seven-layer value stack serves as a framework to understand AI implementation better. It consists of foundational layers involving process design, governance, and knowledge architecture, followed by layers for human judgment, feedback loops, and scaffolding, with the AI model sitting on top. Unfortunately, many companies jump straight to the model layer, neglecting the essential groundwork needed to support it. This oversight is reflected in recent surveys indicating that while many organizations have strategies for AI, a significant number feel unprepared to execute them effectively.
What it means
A real-world example illustrates this issue: an organization using an AI model for support ticket triage discovered that 30% of tickets were misrouted due to outdated classification systems. The solution wasn't a more advanced model but rather revising the underlying processes and knowledge architecture. This demonstrates that even effective AI models can fail without proper support structures. Additionally, as companies like Apple show, merely swapping models without addressing the surrounding framework means that the competitive advantage lies not in the model itself, but in the systems built around it. For organizations looking to implement AI effectively, understanding and addressing these foundational layers is crucial for long-term success.
