joshbjon.es

I’m running an experiment this year. Anytime there is something that is simple and might be delegated, I see if it can be automated by tech instead of given to an intern.

I’ve been hesitant to do this over the past year due to a value for developing others, so i’m also looking for new ways to intro others into a career that involved more valuable and less busy work.

My start on it was a report I get daily. We’re in the O365 platform so I used Power Automate to
1. drop the pdf to Sharepoint
2. archive the email
3. when a new pdf shows up use a trained ai model to pull data
4. add the data to an excel doc

It’s a daily 5 minute task that won’t have to be repeated by anyone ever again. The way I figure it I’m saving about $140 per month on that one task.

AMA.

I’ve been thinking about AI a little more lately after some recent conversations. Since OpenAI launched ChatpGPT, it’s arguably taken more investment dollars relative to the return than any other investment vertical in history.

There’s a lot of buzz around it. Buzz is fine and value is better. I’ve spent the past several months working within a few AI driven tools and here’s the thing i’ve found. What’s here today, is not tomorrow’s answer. This is not a statement about the quality of the tools, it’s an assessment of the stage of the industry. We’re still incredibly early.

Let’s take co-pilot in office as the example. It’s fine and it’s limited. If I want to write a formula faster then that’s great. It does that. Honestly that’s not where most of my brain energy is spent though. Email, it’s kind of terrible. Doesn’t have the context I do for what I’m writing to the point that my feedback to Microsoft one night was that “A monkey on cocaine could write an email better than this.” Really I was missing the point. It’s a Co-Pilot, it’s not going to do my job for me yet. At the same time a human co-pilot would be better.

Predictively guessing at what I want to say next has been a thing for years that Google already gave us. When you think about how you work, that feature isn’t really that helpful.

So, as i’ve had opportunity to use it, the better value has been found in applying that to automated tasks I don’t like. Stuff like:
-capture the data from this pdf report into excel every afternoon
-find empty time on my calendar and preschedule this
-give me the tasks from this meeting

Writing my email for me though? That’s a ways off. So the question is, what’s the IRR we’re looking for on this money being thrown here and how can you capitalize on the models beyond basic use of a Chatbot GPT?