Why Data Strategy Beats AI Strategy

Every organization seems to need an AI strategy. And everyone wants their own chatbot. I think we may be solving the wrong problem. Consider Antagning.se. Finding courses relevant to me takes work, so the obvious AI solution is an assistant that learns about my education, career, interests and plans. But why should Antagning.se need to know any of that? They know which courses exist. I know what I want. My AI can connect the two.

Instead of building another AI, Antagning.se could make its course catalogue easily available as clean, documented and structured data. My personal AI could fetch it and evaluate it against what it already knows about me. Antagning.se wouldn’t need my academic history, career plans or preferences. This separation could apply almost everywhere. Organizations hold domain knowledge. Users hold personal context. Why should personal context be processed by the same system that provides the domain knowledge? Particularly from a privacy perspective, the reverse makes a lot more sense.

Policy

Rather than every public authority separately procuring or developing its own AI assistant, governments could require public authorities to make suitable public data available. In Sweden, for example, the government could make machine accessibility a normal requirement for myndigheter. Let JSON be the rule, and PDF the exception to the rule. Additionally, common standards for schemas, identifiers, metadata and discovery could make information from different authorities usable together.

Privacy

The web was built for humans using browsers, but our new users are machines. They don’t need navigation menus and graphical search interfaces. They need the information underneath. Initiatives such as llms.txt point toward machine discovery, but the particular standard matters less than the principle. Stop making machines reverse-engineer websites designed for human eyes. Give them the data directly.

There is a privacy argument here. Personalized AI does not have to mean that every organization learns more about us. It could mean the opposite. My personal AI can know me, while Antagning.se knows courses, the municipality knows municipal services and the university knows its programs. Intelligence can sit with the user while organizations concentrate on making their own knowledge accessible.

Megadeath

In his inaugural speech in 1961, Kennedy famously said, “Ask not what your country can do for you – ask what you can do for your country.” In 1990, Megadeth flipped the script: “Don’t ask what you can do for your country, ask what your country can do for you.”

In relation to AI, everyone seems to be following Megadeth’s advice instead of Kennedy’s. They keep asking what AI can do for their organization, when they should be asking what their organization can do for AI. Structured and accessible data is going to do more for humankind than any number of AI chatbots ever will.

This text is dedicated to the metalheads.