Software-defined Automation: Obsolete

Software-define automations help us perform tasks based on explicit instructions. Generative AI lets us do that implicitly.

I recently shared this example on Linkedin. It seemed innocuous at the time, but as it percolated and a few readers contacted me directly about it, I realized the impact that implicit instructions might have on a substantial software segment — software-defined automation.

A Simple Gmail Query using Bard and My Email Archive

As I shared …

I'm impressed with Bard's new ability to access my email history, which goes back to April Fools Day 2004 (the launch of Gmail). With a simple prompt, I can generate a summation of all conversations with a specific person. It also responds to queries based on content, dates, To, From, CCs, and BCCs, and even the subject of a thread.

Everything about this example was possible before generative AI. You could search email archives, extract similar hits, analyze the messages with your own eyes, and select the ones that matter.

How is this different from a database query that [smartly] filters your messages?

It’s not.

When we see our personal data (such as an email archive) comingled with a system capable of predicting the next word based on a given context (i.e., your data), we panic. Why is that?¹

The sensation I get from this example is magical automation, not fear. The output in this example is not easily attained without generative AI, but it is attainable. In fact, we could automate this process with code or a no-code platform like Zapier. Several software products exist to digest your email archive and do exactly this.

But this [AI] approach basically says …

All those other ways of doing this will soon be obsolete.