Does Performance Even Matter?

This week while attending Scott Moss’ Build an AI Agent from Scratch v2 workshop at Frontend Masters, we spent a little bit of time talking about Agentic Workflows. I was especially interested because I recently gave a presentation on Workflows in Sitecore, and I have a 3-part companion blog post.

We talked a bit about how you could set up your agent to run through some of the tedious validation steps, like spellchecking, searching for alt-text, looking for HTML errors… and even some things that haven’t been automated for years.

Sitecore already offers a number of built-in actions that will validate your pages when you submit for approval, and I imagine most platforms have similar systems in place. Comparing the processing power it takes for a simple action to ping you if alt text is missing versus an AI Agent analyzing the page to do the same thing kinda makes my head spin.

Does it matter?

The companies that are implementing AI Agents and Agentic Workflows aren’t exactly strapped for cash. And if you look at the history of computing, it’s easy to assume that processing power is going to continue the growth it’s seen for decades.

When we look back in 20 years, is hand-wringing over AI processing going to be looked at like people worrying about kilobytes of data storage?

Or are all the companies relying on agents in for a rude awakening when the AI overlords decide they need to start making money?