Documents, Templates and AI: Oh My!

I don't mean templates for your AI deployment, personal assistants, or Internet of Things thing. I mean templates of documents, sales sheets, statements of work, and so on. You know, those things we spend hours of our lives looking for. Usually mythical materials created by some nebulous force in the org that’s used in recurring situations; those valuable items that you quest through SharePoint, Google Drive, or some file share in the closet, only to find the titles right but it’s from the 1900’s. I often talk about data preparation, data quality, and data understanding as part of the process for deploying ML or AI solutions. It's like step 1B (while 1A is ethics of course).

Step 0 is “Do the things we need even exist?”

Who wants to go adventuring for the Golden Fleece and come back with itchy blanket?

 We should ensure that those high-impact, high-value documents that are going to be searched for exist in the first place, and/or are up to date. ***GASP***

 I feel like we're missing a step not only where is it but also do you even have it? Do you even have templated language in a document that people can back into? Do you already have the current version ready to go? Do you have that current year PowerPoint template everybody's looking for you? DO YOU HAVE THOSE FONTS UPLOADED FOR ME TO GET THIS PDF TO LOOK RIGHT?!?! 

 OOOooooooOOOorrrrr are we just going to go “Yeah, data’s labeled properly and it’s up there.” Then I'm going to go search for and find me the latest company PowerPoint I got one from 2020 and three from 2 other departments circa 2022 and one that’s 2023 close but I’ll have to gut and rework the whole thing.

 What I'm saying in a nutshell is it's not about the quantity of data available, it's also about the quality and for most people the quality isn't only relevance it’s the freshness of the content.

Meaning, is it current? That 2020 Corp PowerPoint template isn’t doing me me any good if I'm living my life in 2024.

I think we're missing a step. Step one information needs to exist. Do you really need to put everything up there? Ask yourself, “Is the content up there relevant in both topic and time, is it current? In no, generate correct artifact, if yes, send it. Are they “daily driver artifacts”, files and information needed for individuals to perform day-to-day job functions, if yes, those go up as well. If enough people, ask for it then yes. I'm going to get crazy with it here, maybe send out a poll or two company wide prior to launching and see what high value documents are needed. Ensuring those are part of the first wave of information to go up.

 Let's stop treating our necessary documents like elusive treasures and start seeing them for what they really are: essential tools for our everyday work. Let's ensure they're not just available but also relevant and up to date, just like how an antique map won't help us navigate modern terrain dated artifacts lead us down bunny trails of extraneous work.

Before diving into the latest AI deployment, let's take a moment to ask ourselves if we have what we need, and if it's ready for use. The true Golden Fleece of our work can be obtained – it's having the right data and artifacts at the right time, no quest required. By leveraging fresh data through implementations like retrieval augmented generation (RAG), we can improve the efficacy of large swaths of the workforce and save lots of consternation and time.