There is incessant discussion of AI that is accompanied with a plethora of use cases that tout cost reduction (time saving), creative outputs, and automation of tasks. Clearly there is value creation taking place. However, how much of this value is being captured by the focal firm, and how much is being dissipated in competition due to the ubiquity of AI? Specifically, what is the basis of AI’s competitive advantage? This question has been asked before about other general purpose technologies, like the Internet. The answer is usually somewhere in the cultivation of a unique interaction space between the technology and the company using it.
For AI, I see two potential sources of competitive advantage: (1) the trained models (the models often include billions of parameters, and the data that is used to train the model by setting the weights of the parameters) and (2) the use of the model for inference (to automate tasks or create outputs). Each of these has unique considerations for competitive advantage.
On (1), the model and data level, there are model makers, shapers and takers. The makers build and train their models. Early makers (like OpenAI’s ChatGPT, Google’s Gemini) with deep pockets offer foundational models trained on massive corpus of data. While currently, the foundational models quite similar, with some minor differentiation that may be visible to users (usually behind a paywall). This differentiation could be based on how the models are trained with slight specialization, or better reasoning or agentic capabilities. However, if all the models are trained on the same (Internet) data that is hitting limits on scalability, and if synthetic data (generated by the model) is largely derivative, there might be a limit to such differentiation among makers. Smaller companies that do not have deep pockets can leverage open source models, but here the models themselves are open and undifferentiated. Shapers can build on foundational or open source models, fine tuning them with their own proprietary data which could create customized AI offerings, that is a differentiator. The lowest but the numerically largest tier, takers, simply use foundational models and their differentiation could be based on how effectively they use the models.
On (2), perhaps the bigger differentiator is the way the AI is deployed and used. Firms that simply offer tool access (e.g., a paid subscription to GPT) to their employees are not really differentiating themselves – although some employees may leverage the tools innovatively, creating value for themselves and the organization. But tool access is low hanging fruit available to all companies. Firms that can train their AI tools (or the foundational or open source models) with proprietary data can create personalized services. These can create competitive advantage, but the deeper differentiator is when companies integrate these tools into core business processes in ways that create data network effects. This automation of self reinforcing feedback loops (where the model refines itself based on success) is the avenue for sustainable competitive advantage.
Due to the substantial resource requirements, large corporations have traditionally led in developing advanced AI models. However, the rise of open-source AI is democratizing access sophisticated technologies. This shift enables smaller companies to fine-tune existing models for specific applications, potentially leveling the playing field. However, achieving competitive advantage with generative AI requires more than access; it demands building organizational and technological capabilities to innovate, deploy, and improve solutions at scale.
The quality and relevance of data used to train AI models are critical. Foundational models trained on vast internet data offer broad capabilities but may lack specificity. Companies that harness proprietary or domain-specific data can fine-tune AI models to better align with their unique needs, enhancing performance and differentiation. AI enables midsize companies to provide personalized experiences by analyzing data to understand consumer behavior and preferences, thereby improving customer satisfaction and loyalty.
It’s evident that generative AI systems can provide companies with significant competitive advantage. Currently, the hype is on the “wow” factor of rapidly advancing AI tools. However, value creation is not value capture if everyone is doing it. The firm competitive edge will not be based on access or isolated use cases, but the depth of integration of these tools into organizational data and business processes.
Leave a Reply