AI Loves IA: Why Information Architecture Still Matters in the Age of Copilot

All Content Left Behind

In Episode 22 of the Guardians of M365 Governance monthly webcast, Ragnar Heil (@ragnarh), Joy Apple (@JoyOfSharePoint), and I welcome guest Sue Hanley—renowned information architect and Microsoft MVP—to explore the evolution of Information Architecture (IA) in the Microsoft ecosystem and how it’s being reshaped by AI and Copilot.

And what did we learn? That Hillary Clinton and Marie Kondo have more in common than you’d think.

The Librarian’s Revenge

Sue’s origin story reads like a superhero comic for data nerds. From data modeler to managing actual librarians (the kind with real books!) to becoming the undisputed queen of information architecture. Her secret? Understanding that organizing content isn’t about technology—it’s about people and process. The tech just follows along.

“You can have a great information architecture and completely destroy it with lousy governance,” Sue reminds us. And in the era of AI? That statement hits different.

Because here’s the uncomfortable truth: AI doesn’t fix your messy content. It just finds it faster.

Search Doesn’t Suck. Your Content Does.

Remember when search was supposed to save us all? Just throw everything in SharePoint and search will magically find it! Yeah, about that…

Now we’re hearing the same song, different verse with AI. “Copilot will just find it!” Sure, if you want it to return 47 versions of “final_FINAL_thistime_Imeanit.docx” and have no idea which one is actually current.

The hard truth? If you haven’t cleaned your content before the AI era, you’re basically paying Copilot to reason over garbage. Crappy in, crappy out—just with fancier technology.

The Marie Kondo Method (Enterprise Edition)

Sue dropped her golden rule for migrations: All content left behind.

Nobody comes over to your shiny new environment unless they’ve earned the right to be there. Think of it as the Apple upgrade model for your intranet. Clean slate. Fresh start. No digital hoarding allowed.

And for that legacy content you “absolutely must keep”? Put it below the line. Turn off search. If you need it so badly, you can manually hunt for it. Spoiler alert: You probably won’t.

Metadata: Alive and Kicking

Here’s where it gets interesting. Some folks think AI means we don’t need metadata anymore. Wrong.

Copilot isn’t using metadata as effectively as it could today, but it will. It has to. RAG (Retrieval-Augmented Generation) depends on metadata to pre-query before hitting the LLM. Without it, you’re just hoping the AI gods smile upon your search.

But—and this is crucial—we need just the right amount of metadata. Not 50 tags per document. Not 3,000 items in a term set (yes, Sue’s seen that nightmare). Just enough to filter and find what matters.

Think of metadata as the grammar of your files. You wouldn’t skip grammar in a document, right? Same principle.

The 20-Container Rule

Want your mind blown? Current SharePoint agents are limited to 20 containers. That means how you structure your content directly impacts what your agents can do.

Sue’s recommendation for policies has always been: store them centrally. One source of truth. Easier to manage, easier to find, easier to keep current. Now add AI to the mix, and suddenly that architecture choice determines whether you can build an agent that answers “Can I bring my snake to work?”

Size matters. Structure matters. IA matters.

Start With What Matters

Sue’s closing advice? Don’t start with the tech. Start with the business problem.

What content is critical to your organization? For most intranets, it’s policies—the “how should I be doing things?” content. For consulting firms, maybe it’s proposals and win rates.

Focus your calories on what drives the business. Get that architecture right. Clean that content up. Then let the technology work its magic.

Because at the end of the day, AI loves IA. And if you get your IA right, your AI will thank you.

The Bottom Line

We’re in a transition phase. V1 of SharePoint knowledge agents. Baby steps toward enterprise readiness. But the fundamentals haven’t changed:

• Clean content beats organized garbage every time
• Metadata enriches files—use it wisely
• Structure your content for humans first, AI second
• Governance isn’t optional (it never was)
• People, process, technology—in that order

Next time you’re tempted to think AI will solve all your content chaos? Remember Sue’s wisdom: You can’t reason over 25,000 documents with broken governance. But give humans manageable containers of content they’re responsible for, and magic happens.

Now go forth and Marie Kondo your SharePoint. Does this content spark joy? No? Then ask it: Are you relevant? Do I need you? Do you deserve to be here?

The era of AI demands better answers.

Ready to implement proper AI agent governance? Contact me, Ragnar Heil, for a consultation on Microsoft Purview and Agent 365 deployment strategies tailored to your organization’s needs. Find my calendar here at our HanseVision Governance Landing Page. 

The post AI Loves IA: Why Information Architecture Still Matters in the Age of Copilot first appeared on Ragnar Heil (MVP): Empowering M365 with AI.

Original Post https://ragnarheil.de/ai-loves-ia-why-information-architecture-still-matters-in-the-age-of-copilot/