Satya Nadella: Businesses that Rely on Only One AI Risk Not Surviving

Microsoft Chief Executive Officer (CEO) Satya Nadella (photo: x @genbeta)

JAKARTA - Microsoft Chief Executive Officer (CEO) Satya Nadella warns companies not to rely entirely on one artificial intelligence (AI) model. According to him, this strategy can make businesses lose control over data, work processes, and innovation capabilities, which ultimately risk threatening the company's sustainability.

In an interview with the Fareed Zakaria GPS program aired on CNN, Nadella reiterated the warning he had previously made earlier this month. He said that companies that submit all data, interaction history, to commands (prompts) to one AI provider have essentially "delegated their way of thinking."

Nadella assessed that companies need to maintain full control over metadata whenever they interact with AI models. According to him, information such as operational history, context, memory, and programming tools should be separated from AI models so that they remain company assets.

With this approach, companies can use the data they have to refine open-weight AI models or build their own AI systems in the future.

He also suggested that companies are not tied to programming tools provided directly by AI model providers, such as OpenAI's ChatGPT Codex and Anthropic's Claude Code.

As an alternative, Nadella encourages the use of an independent layer called an AI gateway. According to him, this approach allows companies to use various AI models as needed without relying on one provider.

"If at some point one of the models is no longer available, the company still has full control over the system and its workflow," said Nadella.

In addition to the issue of dependence on vendors (vendor lock-in), Nadella also highlighted the risk of competition that arises when companies share internal business processes with AI providers.

According to him, when a company integrates AI agents deeply into its operations, AI model providers have the potential to understand how the business works. This condition is considered to open up opportunities for AI providers to provide similar services that compete directly.

This view is in line with concerns previously expressed by venture capital investor Jason Calacanis. In May, Calacanis reminded the founders of start-up companies that were members of Y Combinator to be careful about receiving free AI usage credits from OpenAI.

He assessed that the use of these facilities has the potential to provide an opportunity for AI providers to study the products being developed by start-up companies, then present similar features directly on their platforms.

However, Nadella differentiates the approach between corporate customers and individual users. When asked about the protection of personal data for consumers, he said that sharing data is part of a common exchange of value in free digital services, similar to the advertising-supported business model of the internet.

On the other hand, Microsoft also has business interests in the AI ecosystem. In addition to being an investor in OpenAI and Anthropic, the company provides cloud computing infrastructure services as well as AI gateways that companies can use to manage various AI models.

Despite having commercial interests, Nadella believes flexibility in choosing an AI model will be an important factor for companies amid the increasing use of open-weight AI models that are considered more efficient while giving more control over internal data and workflows.