Your AI Tool Stack Is Now a Values Decision

by Lisa Peyton, full-time faculty at the University of Oregon School of Journalism and Communication, AI marketing practitioner, and Forbes contributor

I came back from Yellowstone last month (I am NOT a camper, and I still fell in love with it) to find that two of the big AI labs had each shipped a new frontier model while I was gone. So on Friday morning I did what I always do. I opened the Zoom room for our monthly Advanced AI Meetup, pulled up the leaderboards, and asked a room full of working marketers what they’re defaulting to right now.

You already know how to make this kind of call. You pick tools, set budgets, and decide how deep a project needs to go every single week. Choosing a model and an effort setting is the same judgment, pointed at a new set of dials. Here’s what I shared, and what the room shared back.

The Leaderboards Moved. Again.

I use Artificial Analysis as my starting place for raw intelligence. In August the top of the Intelligence Index was Claude Opus 5, with Claude Fable 5 right behind it (and in my own Opus 5 vs Fable 5 bake-off, the leaderboard favorite lost). As of this month, Claude Fable 5.1 sits at the top, tied with GPT-6 Astra.

Here’s the catch. For writing, the boards don’t agree with each other. One creative writing board has Fable on top. Another moved GPT-6 Astra to number one in September. And one of my co-hosts made the point I’d underline twice: she still prefers ChatGPT for true creative brainstorming, and no leaderboard is going to talk her out of what she sees in her own work.

That’s exactly right. Leaderboards are a starting place. Your own test is the answer.

My Test: One Newsletter, Two Models, One Red Pen

My standing test is my Weekly AI Marketing Brief, because I run it every week and I know what good looks like. It has two stages. First a news editor stage, where the model goes out, gathers, and reasons about what deserves a spot. Then the writing stage.

I ran Claude Fable 5.1 and GPT-6 Astra through both with identical prompts. Fable took it, 42 stars to 31 out of a possible 55 on my scorecard. The full breakdown and every receipt is in the Fable vs Astra newsletter test.

So Fable 5.1 is still my default, with a few exceptions. When I need titles, a quick punchy creative something, or personas, I still go to ChatGPT.

The Finding That Surprised Me

The labs are not making this easier on us. We used to choose a model. Now we choose a model AND an effort level, and more effort means more tokens.

So I tested that too, and Fable 5.1 on MEDIUM wrote in my voice better than Fable 5.1 on high. (The effort settings get a full workout in my Fable 5.1 vs Fable 5 writing test.)

Here’s the working rule I landed on:

  • Medium is my default writer. Anything voice-forward, anything where my rules are already loaded in a skill file, any quick question.
  • High is my researcher. Client strategy, comprehensive research, anything where I want the model to cast a very wide net and go deep.

That “rules already loaded in a skill file” part is doing a lot of the work. If you want to build skill files of your own, The Ultimate Guide to Meta-Prompting Claude Skills for Marketers walks you through how I write mine.

Your work is different from mine, so your results will be too. Pick one task you do every week and make it your standing test for every new model.

What Does All This Cost?

Someone asked how many tokens my test burned. Honest answer: I’m as good at tracking tokens as I am at balancing a checkbook. What I have done is price out strategic work on Fable, and one full strategy run came in under a dollar. When a client is paying thousands for that strategy, a dollar of tokens is a rounding error.

The other end of the spectrum is the daily reports I run for our Catalyst Slack community. Five reports a day at a dollar each does not scale. I had Claude rework them to run in off-peak hours on a smaller model, and they now cost pennies a day.

Context is the whole game. Spend on judgment. Economize on repetition.

The News: Docs, Slides, and an Ad Platform

Anthropic launched Claude Docs and Claude Slides. My co-host built her deck in Slides and loved editing right in the pane and exporting without ever opening PowerPoint. My own first brush with Docs was rougher. It changed my defaults mid-task and I wanted my simple Markdown file back. Test it against your own workflow before you switch.

Meanwhile OpenAI announced sponsored agents, which let an advertiser host a branded chat experience inside ChatGPT. I’ve said before that OpenAI is becoming the ad platform and Claude is the work platform, and this only sharpens that. I don’t trust an AI assistant whose business model is selling ads. I also think the offer is compelling, and if your buyers are using ChatGPT to find things, you may need to test it. Both things are true.

Your Tool Stack Is Now a Values Decision

The biggest conversation of the morning was about Pacing the Frontier: the open statement signed by nearly 1,400 employees of frontier AI labs, Dario Amodei’s essay calling for embedded evaluators and democratic and global coordination, and Anthropic’s new report saying Claude now leads 26% of its own model R&D work.

The room was skeptical, and these are seasoned marketers who know PR when they see it. I hear that. Here’s where I landed.

I can’t turn AI off. I couldn’t teach what I teach or make a living without it. At the same time I’m asking big questions about the technology, and holding those two things apart every day is its own kind of tiring. (If that tiredness sounds familiar, I pulled together nine evidence-backed practices for AI burnout last month.) Who knew the tool stack sitting in a marketing department would become a values decision? It has.

What I can control is where my time and money go, and what I let into my feed. That’s the practice. Choose on purpose, curate hard, and teach the people coming up behind us to do the same.

Come Think Out Loud With Us

These conversations are why I hold this space every month. It’s small by design, the chat is lively, and nobody has to have it figured out.

Hit me up on LinkedIn and tell me your default model and effort setting. I would LOVE to compare notes.

Made with my team of AI superheroes and my own skills, including AI-generated imagery. Every opinion, edit, and fact-check is mine. I am an AI practitioner, professor, and pioneer helping marketers put AI to work with purpose. Find more resources at lisapeyton.com/ai-marketing-resources or connect with me on LinkedIn.

Similar Posts