Competition against structured content

Companies want to hear that AI will automate all the things and therefore, it’s going to be So Easy. But unfortunately, we have the Iron Law of Life:

YOLO = GIGO

It’s worth turning, once again, to the history of car manufacturing and particularly, the transition from custom-built cars to mass production on an assembly line. (Note: Tony Self wrote about this back in 2012!)

We have a couple of truisms in automation. One is the idea of 10x productivity. Basically, when you automate, you get 10 times the productivity of when you do things by hand.

First, we only had custom-built cars. 

Then, along came Ford and the Model T, which was famously not a great car, but had the enormous advantage of being cheap. That opened up a mass market because far more people could afford to buy the cheap car.

Now, let’s consider how we got from the Model T to today’s car production process. Over time, we developed more cars and simultaneously got more efficient at building cars. This was the result of:

• Common platforms (for example, the Honda Odyssey minivan is built on the same platform as the Accord sedan). The starting point for two different models is the same, but the end result is quite different.
• Standardization of parts, especially “invisible” parts. Car companies don’t produce their own tires; they buy them from a tire manufacturer. Although cars come with “standard” tires, you can replace them with third-party alternatives.
• Automation and robots. Assembly lines make extensive use of automation and human labor. A robot might move a heavy part, but the human makes sure the part is properly aligned on the chassis.
• Fault tolerance. When you are custom-building a car, you can accommodate slight deviations from part to part. In an assembly line, you need consistency to ensure parts fit together, so the tolerance for deviations decreases. 

Now, apply these concepts to the content production process.

• Common platforms: You can create content variants from a single starting point.
• Standardization: You need external data and content sources to fit into your production processes seamlessly.
• Automation: You can use automation for quality control and to do the heavy lifting (extracting the right details from a sea of data maybe?)
• Fault tolerance: You need tighter content to make sure that all your automated formatting and publishing workflows perform as expected.

Just as with automotive assembly lines, we need rigor and predictability in our digital content supply chains. 

As consistency and semantic value increases, so does the productivity of your content production process. Consider the content development process levels, which I wrote about more than 10 years ago (!!):

1. Crap on a page. There is no consistency in content. For example, two white papers from the same company are formatted inconsistently, are often badly wwritten, and do not use consistent terminology. Two audio files might be encoded differently or have wildly varying levels of audio quality.
2. Design consistency. Content appearance is consistent, but the methods used to achieve the look and feel vary. For example, two HTML files might render the same way in a browser, but one uses a CSS file and the other uses local overrides.
3. Template-based content. Content appearance is consistent, and the methods used to achieve the look and feel are consistent. For example, all HTML files use a common CSS file, or page layout files use the same formatting template. Graphics are created, scaled, and rendered the same way.
4. Structured content. Content is validated against a template by the software. This usually means that XML is the underlying file format. Information is organized in predictable, consistent ways.
5. Content in a database. Information is stored in a database and can be searched and manipulated in interesting ways.

Read more in our blog post, Why is content strategy implementation hard?

If you want to maximize automation, you have to have consistent input. 

AI content tools are like robots in the factory. They work best if the input is predictable and consistent. If you want 10x productivity improvements in large-scale content operations, you need structured, semantic content.