AI capabilities are advancing incredibly fast right now, and I think a lot of businesses are wondering what they should be doing to prepare. So I just want to talk about one thing that I think every business should be doing right now, something we do for all of our clients, and that's building a knowledge base.
The thing to realise is that no matter how capable these AI models become, if you wanna fully take advantage of the efficiency that AI can bring, they need to have access to the information that matters for your business. Like, all of it. As AI gets more and more capable, you're eventually going to want to integrate it into every part of your business.
And therefore, anything that's important to even one person in your business is important for the AI to have access to.
So let's talk about how you want to structure that information and how we collect and organise it for our clients. There's still a lot we're learning about the best way to organise this, but there's a few things we can be confident about. We can be pretty confident that you want to store the information in a series of markdown files.
This is AI's preferred file format. And you wanna make it as easy as possible for the AI to know which files it's going to need for a given task.
Eventually, this knowledge base is gonna be very large. You'll have dozens, if not hundreds of files, and some of them might be dozens or hundreds of pages long. Because the AI can't pull all that context into every task.
It needs to be very clear which files it should use for which task. Each file should have a clear descriptive name and then a short description that explains exactly what it is and when it should be used.
I also recommend strictly separating documents that contain internal information versus documents that you would gladly share publicly. Presumably, this AI is going to be drafting public facing messaging and content. You don't want internal information leaking in.
Next, you need a place to store and organize these files. Currently, we create a Claude project for each of our clients. We call it their brand ambassador.
Each of the markdown documents is stored in a Google DOC and then uploaded to the project files. Storing them in a Google DOC means we can use the same documents in a Google Notebook or a ChatGPT project and don't have to maintain multiple versions of the documents.
We have about 12 standard documents that we use for every client. These are things like a company overview, a complete description of every product and service, content guidelines, a list of every important page on their website, a competitor analysis, a brand voice and style guide, and then anything else that's important for the AI to know about their business.
And then in the project instructions, we again detail every document, what it contains and when it should be used. That way, the AI always knows which documents it needs to retrieve for a given task. I do wanna mention a lot of people are storing these kind of documents locally and there are some real advantages to that.
You can use something like Obsidian to organise all the documents and link them together. Then if you're using something like the Cloud Desktop app, you can even have AI edit the files for you.
This can work well for very small teams, but if there are more than two or three of you, it can become a version control nightmare. I'm sure better solutions are coming soon for enterprise, but having the documents inside of a cloud project right now is the best solution we found.
Point, though, is that it's the documents themselves that are gonna be valuable. However this evolves in the next few years, spending time making these documents is well worth the effort. Now, having an AI that fully understands every aspect of your business and website and brand is a huge unlock.
It allows AI to do hundreds of tasks that it otherwise would be incapable of.
