I’m a Professional Writer Who Uses a Very Controversial Tool. It’s Not As Scary As I Thought.

Sign up for the Slatest to get the most insightful analysis, criticism, and advice out there, delivered to your inbox daily.

A.I. didn’t write any part of this story. A.I. didn’t outline it for me and didn’t catch typos on a first draft, though I don’t have any more of a moral issue with that practice than I do with using spellcheck. (Slate happens to have a terrific copy desk, a perk not available at every media outlet these days.)

There’s still a chance I wouldn’t have written this story without A.I. That’s not because the tech is doing anything a person couldn’t do, but because it has gotten good at removing friction in the spots where there was once just enough of it that I used to make problems for myself.

I prefer to be skeptical about A.I. because people have already found so many destructive uses for it and I don’t trust Sam Altman or his competitors as far as I can throw them. I’m also aiming to be open-minded these days, because it’s just not true anymore that the tech isn’t impressive. When I talked with the prominent A.I. critic Ed Zitron for Slate in February 2025, I agreed with his view that chatbot apps had generated throngs of users but were still lacking in use cases that would take them from a cool toy to a useful tool for anyone who wasn’t spending hours trying to learn how to become a power user. I now think that view is impossible to hold. Spend a few hours in Claude Code or Codex. Tell it to interview you about what you want and, after about five minutes, spin up a portfolio site based on your answers, and then think about what actual trained developers might be doing with that tool. These things do not have to rewire every element of day-to-day life to be unbelievable.

If you’d have asked me a year ago how I was using A.I. in my writing process, the answer would’ve been “Not at all, and you’re a nerd for asking.” Now the answer is different: I use it most days, although not in the ways people talk about when they’re dreaming up a future in which technology will either replace my job as a journalist and creative professional or make it a lot less stressful.

I see a lot of Instagram Reels from influencers who insist that Claude now runs their lives, or that they “had Claude redo my entire social media, and it got me 50,000 followers in three months.” I suspect that almost all of these people are being fantastical, or doing what is sometimes called “lying.” Claims that the latest Anthropic release “just killed graphic designers” or “just killed accountants” or “just killed social media managers” all land as BS. I find that the most worthwhile uses of the tools in my field are a lot less impactful than that but still valuable.

The most frequent use of generative A.I. in my workflow is, from my perspective, not generative at all. I make a lot of phone calls for stories, and meticulously transcribing those interviews has been the bane of my existence since I was in high school. A.I. transcription has gotten extremely good in the past few years and has taken a notable step forward over the past several months. I sometimes use the transcription platform Rev, which offers an A.I. transcript you can scrub through along with the audio. I’ve also run transcripts on a local model on my computer, which does a good enough job for shorter conversations.

The ability to upload a transcript to Claude, then instruct the bot to “pull anything this person said about this issue, including if they came back to it later,” is quite nice when working with transcripts of more than an hour or two. Anyone too lazy to check the parts of these transcripts that they use in a story is asking for trouble. (I find that both Claude and Codex make things up way less often than they did a year ago, though the frequency is still not zero.) But having a bot read a transcript and identify the relevant sections for me to verify is a massive time-saver, one of my favorite technological evolutions of any product I’ve used in my career.

Along the same lines, dictation software has made a massive leap this year. Products like Wispr Flow and Monologue (I use the latter app) are an apt use for the technology that undergirds A.I. platforms. (Even a critic would admit that the bots are good at guessing sequences of words.) I find it amazing for quickly capturing thoughts and outlining things for myself while I’m on the move. I like that I can now record a voice memo while on a nice neighborhood walk and tell my phone to put the transcript somewhere I know I’ll find it later. Walking and typing is dangerous, but walking and talking like a maniac into my own recorder is merely antisocial. For a maladjusted writer, that’s progress.

A significant number of my recent stories at Slate have come from a Notion database I call “Idea Dump,” the contents of which are mostly derived from me talking into my phone and telling an app to send those thoughts to my Notion system. I’ve made an A.I. “skill” (just a Markdown text file) that ensures that these ideas get the right labeling so I can easily sift through them when I’m back in front of a big-boy monitor.

None of this has changed the parts of my work that I actually liked before or felt were valuable to my process to slog through. I’m not looking for a bot to remove the grist of plotting out and writing something good. I’m looking for A.I. to help me jot down and store my thoughts in the right place when I’m not in a position to easily do that.

I like this workflow because it addresses what I’ve long understood to be one of my weaknesses: thinking of something exciting, forgetting about it or converting it to an unintelligible scribble in my Apple Notes app, and never capitalizing on what felt like a great idea in the moment. Machine-learning dictation and agentic A.I. have alleviated that issue, and the best way I can measure this is by the hundreds of little ideas I’ve got stored and the relatively few nonsensical one-line notes I see in my Notes app. (A.I. has also helped me capture a bunch of awful ideas that sounded better when I was dictating them.)

The tools have been useful in my day-to-day life too. This year, I’ve been trying to lose weight—17 pounds down so far, I’m happy to say—and shifted to using a kanban board for my weekly meal plan. I am unlikely to deviate from a system when I see things laid out in front of me. But I’m notoriously terrible at either forgetting ingredients at the store or duplicating them, so I used Codex to rig up a shopping list that links into my project management system and pulls from the recipes I upload into my meal plan. I do not believe that the entire stock market should rest on my having access to a grocery list power tool. I do like, however, that a 10-minute buildout has reduced how many bags of pasta I have open at any given time. I’m less likely to do multiple laps around the store.

I don’t think I’m so special that I can use the tech without any risk to myself. My anxiety about A.I. is not that it will make my job obsolete but that it will make me a little dumber. With the consent of all involved—a key point, given that I live in a two-party consent state and also like to have the trust of my co-workers—I’ve deployed an A.I. notetaker for a handful of meetings. It’s a nice tool in theory, a way to pay attention to the other people on the call and not my Apple Notes interface. But I’ve found that not having to take notes has had the opposite of the intended effect on my absorption of information. It prompts me to be less attentive, because I perceive the notetaker as a safety net. This is the digital media professional’s version of a high school kid being “deskilled” by using ChatGPT to write their paper.

Sadly, A.I. did not even spare me from spending several hours thinking through and then writing this story. I couldn’t dictate it beyond my initial outline thoughts, because I find that I can’t replicate my best writing voice while speaking into a phone or computer. (This is disappointing, given that my stylistic North Star has always been to write how I’d talk to a decently close friend.) Maybe that’ll change. Maybe not.

These few months of experimentation have left me comfortable that A.I. will not replace human writers at popular publications, no matter how many times an audacious media executive tries to see what they can get away with. Even if the writing is good enough to pass a blind taste test, our speciesmates seem to hate it when they find out they’ve been reading a machine’s work. I agree, and more than that, I wouldn’t have started writing if I didn’t like writing. The most optimistic case I can make for writers and A.I. is that we’ll get to spend more time agonizing with our thoughts, because we don’t have to spend as much time scrubbing back through transcripts at 0.75x speed. You cannot take the pain out of writing, but you can lessen some of the grunt work and redirect your energy, and despair, elsewhere.