Issue 10

In issue 9 of Model Thinking, I introduced the topic of content debt, and I tackled it from the angles of content structure and content strategy. Turns out, it’s a topic that resonated with a lot of people. When I posted about it on my LinkedIn profile, the post got above-average engagement, my subscriber count shot up, and the newsletter issue became my most-visited issue so far.

So, let’s spend some more time talking about content debt. In lieu of most of my typical newsletter sections, I’m going to address feedback to my post or its reshares.

I didn’t know there was a name for the outdated content I deal with.

Yes. In software companies, there’s often talk about technical debt (e.g. bugs, code that could be optimized, etc.) and sometimes about design debt (e.g. design bugs like problems with stylesheets or the wrong typeface or color, user experiences that could be improved, etc.). Content debt just isn’t recognized as widely by leadership because content rarely is viewed as the asset that it is.

It seems to me that any generative practice will have debt. You’re in building construction? Automotive manufacturing? Congrats, you’re creating maintenance debt.

Isn’t content debt inevitable? It seems like a balancing act between risk management and innovation.

Yes, probably so. This is an interesting insight.

I used to manage a weekly newspaper (let me know if I need to explain that 😉). We knew that some content would always feel stale by the time an issue hit the streets. We also knew that some content would trigger discussions and discoveries and lead to changes in our community.

We could have managed the risk of stale content (or content debt) by not publishing anything. But we had to innovate and tell new stories weekly. Sometimes, we’d have to deal with content debt by publishing corrections in the next week’s issue.

So what, exactly, should teams be doing to address content debt?

My initial response to this line of questioning was this:

Now with a little more time to think about this, I also add this (and would probably insert it at the beginning of my list):

Of course, I’ve got a small consultancy that can help organizations with content architecture and content operations, and I’d be happy to talk more if you’ve got a content debt problem to address. You can book a call via www.collinscontent.com.

But can’t we just solve the content debt problem with artificial intelligence (AI)?

No, not entirely. I do think that AI can be part of the solution, but you still need humans in the loop. I’m not the only one. See Jorge Arango here, here, here, and here. Jorge’s smarter and more experienced than I am.

One of the most effective applications of AI to content comes in the content operations section, where AI automates workflows and checks grammar, style, and other governance-related items.

Yes, there’s promise for using AI-adjacent technologies (to avoid getting too technical) to organize, label, and enrich the meaning of new or existing content. I’ll even say that this is something that content professionals should be part of exploring—don’t leave it to the engineers!

However, I caution against thinking that AI is the starting point for solving this problem. There are clear human-led steps to create the foundations of meaningful, useful, valuable content that systems can enhance.

One of my favorite examples comes from a knowledge graph conference talk. The search feature in Google Maps allows you to ask “Where can I buy milk nearby?” and it will suggest the closest convenience stores and supermarkets. This was manually built and even human verified in the early days, but AI-related technologies expanded the information and capabilities of the “content” of Google Maps. (And, it should be noted, it’s not always accurate and there is “content debt” and mechanisms to suggest edits.)

I like this direction and application of AI. But it needs content or knowledge professionals to design and govern it. It also doesn’t address the nitty gritty details of real-world content debt at the scale of many organizations.

Structurally, we should still start with well-structured content models that enable human content creators and editors to create and maintain content blocks easily and which can shortcut the need to train an AI tool or that require heavy computing loads to process unstructured blobs of content.

Things that are bouncing around in my head as I synthesize a range of ideas

John Collins

Thanks for reading! Did someone forward you this email? Subscribe here​ If you’re already a subscriber and you found value in something here, tell your friends and colleagues to subscribe!

Welcome to the 23 new subscribers who have joined us since the last issue of Model Thinking.

(As an Amazon Associate, I earn from qualifying purchases via affiliate links.)