The piece discusses the ongoing debate around whether enterprises should own their AI models or rely on renting them from companies like OpenAI and Anthropic. The author, Manish Sood, emphasizes that the real value lies not in owning the models but in owning the systems and processes that surround these models—the so-called “loop.”
Key points include:
-
Ownership vs. Renting: While some executives advocate for owning AI models for sovereignty, others view renting as a more pragmatic approach.
-
The Loop’s Importance: The loop encompasses proprietary data, governance, workflows, and continuous learning, which are crucial for turning AI outputs into business results. This loop can adapt and gain value over time, unlike static AI models that can quickly become outdated.
-
Costs of Model Ownership: Constructing and maintaining a proprietary AI model is resource-intensive and can lead to escalating costs without guaranteeing superior performance.
-
Data Quality: Even with a robust loop, if the data used is fragmented or unreliable, it can lead to ineffective outcomes. Trustworthy data is essential for the success of any AI initiative.
-
Investment Justification: Companies should consider if they have the resources and infrastructure to operate a loop effectively before committing to building one.
Sood concludes that true value in AI comes from a well-established loop, provided the foundation—high-quality, consistent data—is in place.
