Context is Everything
I keep seeing blog posts and reflections* about how people tried to use AI for something they already like doing, and then getting discouraged because AI doesn't contribute anything to their workflow. I think I'm coming around to identifying how I'm using AI in contrast to how others are.
Many of the reflections dive into the idea of using AI to "increase their output" and translating that as "having the AI do the work for me." Ultimately, every post I've read has ended with the sentiment that they dislike the loss of cognitive processing and would prefer to do it on their own and live without AI. It's probably the best conclusion to come to in that scenario.
My use is a little different. I don't use AI to write. I enjoy writing. Why would I outsource something I enjoy doing? I enjoy the process of organizing thoughts, finding the best way to frame/phrase something, and constructing ideas in a way that other people can understand them. For me, writing is a non-negotiable. I write to process my life and what I'm thinking/feeling and having someone/thing else do that for me is weird. I also love looking back on my writing over time to see how my voice has evolved. I used AI to write once and I hated it. That's not my voice, I'm not putting my name on that. For the same reasons, I wouldn't hire another person to write for me.
I am perhaps using it "wrong" — at work, I'm not automating the small stuff, I'm using it to learn and understand more of the big stuff. And I don't always start with AI. Usually I read a book or a few articles/papers, then come to AI to help consolidate and synthesize my notes/understanding to ensure I grasped the main points correctly, and if not, I do more reading and note-taking. For me, AI goes into a process that already exists, it doesn't create new ones. It helps to get to goals faster, expands my knowledge, and assists with context-driven situations.
At work, I have a few use cases for AI:
- Research the best ways to handle scenario-specific issues. Because we have an LLM that we use internally, I can feed (some) company data and information into it for help. I am limited to my context; this has helped to discover many (better) ways to solution for tasks that impact multiple people or teams.
- Save time on tasks I would not otherwise do. I've accomplished many things in a few weeks that would have otherwise taken several months, mostly due to time: organizing large bodies of information, completing content audits, and data analysis. I still worked on the results; no output is ever perfect and always requires review and correction.
- Boost people. I use it to summarize what "we" (chat and I) did using different perspectives to share with my colleagues. I'll give an example of this in a minute.
- Knowledge Agent creation, which is still in progress. Rather than "adopting for the sake of adopting" I'm taking a meaningful and strategic approach to organizing the knowledge in a way that the machine can understand it, and which will be easier for me to maintain long-term. So far, it's been useful for me personally, though it's not quite ready to be released to others because it's very incorrect given existing made-for-humans docs (this is another reason for the knowledge audit). I am also exploring ways to create agents that don't rob others of their cognitive processing but rather guide them in perhaps a more interactive way than a document would.
Example Use Case: I'm doing a knowledge audit of all technology docs scoped to my team. After auditing 23 docs (still more to go), I paused and prompted AI to help me identify the types of knowledge that the team requires to do all this work. Just in those docs, and based on my detailed prompting, it revealed 8 overall skills and 10 different types of knowledge required to process the work.
I was also able to get patterns across documentation that provide insight into what the documents expect from the readers. This insight helps with revisions and updates: if the documents expect too much from the readers, maybe they need to be fleshed out more. If they expect too little, maybe that reader-document relationship needs to be evaluated. Since I had a heavy part in writing those guides, this was a huge blindspot. I did my best to write for different needs and follow all the best practices, but the guides still expected people to basically be me.
Then I shared it with the team's manager. This was not information we didn't already know: she knows what skills her team needs. However, having it formally written out with examples and "titles" for each skill will be helpful in onboarding future team members, improving the documents, and letting the team know just how much they really do.
Everyone gets imposter syndrome and in the flow of daily work, you don't often recognize what you're thinking or doing. This kind of perspectives helps people truly understand not only their "value" to the company (which is helpful in performance reviews), but also their personal qualities. Maybe this job brings those types of thinking out, but they are the captains of their own knowledge.
I don't have the training or taxonomy to do that on my own, but I have just enough understanding of how it works to review an output for accuracy. I don't feel like I did any "cognitive offload" to AI — it's not work I would have done otherwise. The knowledge audit would have taken several months and then it would only be used to update the guides with gaps and to ensure currency/accuracy for priority content. By myself — a team of 1 writer/knowledge manager — it would be unrealistic to expect more.
If I ever use AI for things outside of work, it's to manage executive dysfunction. I use it to make sure I eat right, which is something I've been struggling with my entire life. I sometimes use it to pressure test ideas. The rest of the time, it's used to learn how AI works, understand the pitfalls and limitations and biases, and then write about it here.
I'm not trying to imply there's a "correct" way to use AI; how you use it 100% depends on you or your company's needs. However, I think it's interesting the reasons people try it, what they find out about themselves or the technology as a result, and whether they decide to continue experimenting or not. Most of what I've seen (a very small sample of experiences) has been people who try to use it for something they would prefer to do solo, then they stop using it altogether after finding out it's not helping.
On the note of thinking and knowledge types, I'm sure that plays a huge part into how you approach AI and how you experiment with new technologies. I've grown much technology and testing knowledge in my experience working in tech, so my approach to everything tends to be varied and focused on multiple angles/uses, not just one and done.
Context is everything. Ultimately, my journey so far has taught me AI is only as meaningful as the context we bring to it. People are being told to use it as a substitute for their own humanity and the things that drive them, to automate creative, messy, and satisfying acts of thinking/making. When you move away from "outsourcing your mind" and towards "expanding your capabilities," experimentation is completely different.
For me, AI will never become a ghostwriter. It will never draw in my sketchbook for me, and I won't be using it to "think less." Instead, it helps bridge the gaps when executive dysfunction (or being a team of 1) gets in the way. It doesn't replace my cognitive processing, it steps in where my manual bandwidth ends.
The beauty of navigating new technological landscapes is that there is no single "correct" roadmap. How you interact with AI — or whether you interact with it at all — is a deeply personal choice based on your unique cognitive style, needs, and what you value doing ourself.
If you try and hate it, walking away isn't a failure; it's successful boundary setting.
If you use it strictly as a specialized assistant for specific needs, that's a valid integration.
If you look at the landscape and decide not to participate at all, that's a powerful stance on how you want to govern your time and energy.
I think most importantly, remember that we are all experimenting in real-time with something that is still "new." Figuring out what it means to live and work alongside AI without losing ourselves in the process is part of that experiment. There should be no guilt in opting out, and no rigid dogma in opting in. The AI can't dictate how you work or live, but you can dictate how it shows up on your flows.
*I won't link, I don't want them to feel called out. We are exploring a new technology to see what it can do for us, layered with guilt and confusion without regulation. There's some trial/error involved.