What is a company’s most valuable asset in the AI Era? That’s an open-ended question to which different business leaders will have very different answers. Perhaps it’s proprietary data, a literate and knowledge-hungry workforce, or access to a budget for the latest technology rollouts.
According to Microsoft CEO Satya Nadella, in the AI Era, an enterprise’s accumulated knowledge, captured through its interactions with AI, could become one of its most valuable strategic assets. Nadella argues that this knowledge should be controlled and owned by the company. He laid out this argument in a recent thought leadership post on X, and since he is in charge of one of the companies leading the AI Revolution, it’s worth exploring his perspective in a little more depth.
The Reverse Information Paradox
In his essay titled “The Reverse Information Paradox,” Nadella challenges the traditional notion laid out by economist Kenneth Arrow that to convince someone to buy your knowledge, you have to reveal some of it.
Here’s the quote in full: “There is a fundamental paradox in the determination of demand for information; its value for the purchaser is not known until he has the information, but then he has in effect acquired it without cost.”
In the AI Era, says Nadella, this idea is flipped on its head because users, the buyers, take risks by giving away their knowledge to make the best use of the tools they’ve invested in.
“You essentially pay for intelligence twice, once with money, and again with something even more valuable: the proprietary knowledge you must reveal to make that intelligence useful,” says Nadella. “The better you want the model to perform, the more of that knowledge you have to feed it!”
Essentially, every prompt and correction teaches the model something about a business, meaning that a customer is paying for AI while also improving it. In the end, it’s the seller that learns more about the buyer, flipping the information paradox on its head.
The Bigger Concern
Nadella says that companies aren’t just sharing their data, but also sharing institutional intelligence. It’s this intelligence that’s far more valuable than the data itself.
However, Nadella strongly believes that this intelligence should belong to the companies. “In consuming intelligence, you are creating intelligence. And what you create should belong to you,” he says.
Nadella explains that when AI providers collect all this knowledge, it’s they who become smarter, with every company improving their services, but in a way that is totally economically imbalanced, because they don’t get a similar insight into the providers’ own improvements. “It’s imperative that we distribute the learning infrastructure to every firm so that they can control their own learning loop,” says Nadella.
“Patents solve one aspect of Arrow’s paradox,” says Nadella. “They let an inventor disclose an idea without simply giving it away. The Reverse Information Paradox needs its own equivalent.”
His solution is to create a trust boundary for each company. This would regulate its prompts, memory, evaluations, agent traces, fine-tuned models, and institutional knowledge. Nadella outlines five pillars to his idea. The first is control over what the models output. Next, companies should have the capability to train or fine-tune models privately.
He believes that organizations should have the freedom to choose their AI vendor without being locked into a single option, and that costs should be manageable by allowing companies to mix and match models. Finally, by bringing these four pillars together, companies can compound their knowledge and continuously improve their own AI systems.
Final Thoughts
Ultimately, Nadella is advocating for the user. He is envisioning, or perhaps nudging us towards, a future that won’t be dominated by a few large foundation models with all the knowledge.
Instead, he believes that enterprises should create their own AI learning loops built on these models, allowing them to maintain control over their accumulated knowledge and expertise within the organization.
While this is certainly an optimistic, democratic, and inclusive approach, it also aligns with Microsoft’s corporate focus on multi-modal orchestration. But I’d like to give Nadella the recognition he deserves here as a corporate leader putting the customer first.
Ask Cloud Wars AI Agent about this analysis
