It has been one year since I wrote about AI in part 1 of this series. I’d still like to say I don’t do AI, but unlike this time in 2025, I now intentionally spend time working with AI agents.

Shortly after I wrote that post, I was added to a committee at work to study and determine (at least in part) how AI will be used at our institution. This has been incredibly eye-opening and being able to collaborate with colleagues from across offices and divisions leads to many more questions than solutions. In December, I was assigned to take a Google AI Essentials course. I bristled at first, but it turned out to be an incredibly helpful exercise. So helpful in fact that I’ve moved on to Google Prompting Essentials.

These courses are best for me because my workplace is a Google workplace and the only official, sanctioned AI tool available is the Gemini suite of tools. This includes Gemini and Gems, Studio, and Notebook. I began mapping out business processes that I personally manage, estimating the number of hours they require in a week or month, and creating workflows to see which can be handed off to agents. A friend at a tech company advised that to really understand how to work with agentic AI, one should spend about an hour a day on it. I don’t have that time to spend–nor the inclination–but once I had a handful of processes documented (on paper), I set aside a couple of hours a week to see if those processes could be helped, or even automated, by the Gemini tools.

Here’s what I’ve learned:

1. To successfully use AI requires planning and if your task is technical, some knowledge of if-then statements.
2. Learning to write a prompt and then iterate through the results requires critical thinking.
3. 1 and 2 take time and patience, especially if part of your goal is efficiency. (And it should be in order to avoid using resources you don’t have to use.)

And an additional VERY IMPORTANT (or just interesting?) thing I’ve learned is that the platforms themselves are changing and “improving” so quickly–exponentially quickly–that something impossible in one week is possible the next. This results in a feeling that you constantly need to be IN THE AI. Some features that my business processes require hadn’t been developed even a month or two ago and it’s exciting to find that a problem can now be solved. On the other, I wonder if I’m being manipulated somehow to spend more time in the tool. I’m sure I’m not the only person sensitive to that after spending so many years at the mercy of algorithms.

Because I now understand better how to craft a prompt, I’ve found a use for AI in my personal life. The first use is shopping. I’ve used ChatGPT to find glasses instead of going to a dozen different online stores and trying to find frames that match my preferred measurements. The second is tech support for WordPress. After attempting to find information about taxonomies and custom WordPress post types through a traditional search, and reading a fair number of explanations from developers in various corners of the web, I turned to Claude. Claude didn’t solve my issues, but it made me realize I hadn’t completely mapped out my process. I’ve backed away from that until I can do that. I accept that it prompted me to iterate for myself.

I feel less like ranting against AI these days, but in some ways a lot more cautious. While I’m carefully ensuring that I am responsible in my use, I’m watching the fight over data centers all over the country. I’m watching climate change accelerate. I’m accepting that I’m part of the problem, even if I’m being cautious. I’m still incredibly worried about what AI means for education. I still stand against generative AI being used to take the place of human creativity. I still believe training LLMs is the result of theft. It’s all a bit difficult to reconcile in my head, but I’m trying.

Things that give me hope:

1. Credo23 – Founded by Justine Bateman, the C23 badge on a film indicates no generative AI has been used in the project. There is an accompanying Credo23 festival.
2. <RSL Media> – Founded by Cate Blanchett, Nikki Hexum, Doug Leeds, and Eckart Walther, this organization is responsible for a standard that defines “machine-readable AI usage restrictions and licensing terms for creative rights, including music, films, books, identities, characters, and marks.”
3. The kids – There are a lot of young people saying no to AI. I can’t wait to see how they take control of their digital lives.

Feature image information: JPxG, Public domain, via Wikimedia Commons