On the 26th of February I posted on Instagram that I was done with generative AI. Done with the hype, done with the outputs, done with the noise. By the 4th of March I had migrated to Claude. This post is the honest account of what actually happened, what pushed me over the edge, and the very specific way I am now using AI in my business, because it looks nothing like what most people are doing.
Why I Almost Quit AI Entirely
There is a lot of hype in the entrepreneurial space right now about what AI can do for your business. And I get it. The promise is compelling. Faster content, faster systems, faster everything. But when I looked honestly at what I was actually producing using ChatGPT, I was not impressed.
The content was not good. And I say that as someone with 25 years of marketing experience who audits client content every single day. I can see AI content when it lands in front of me. The staccato rhythm, the contrast framing, the way it handles objections in a way that sounds like it is handling objections rather than actually speaking to a human being. You can smell it, even when someone has gone through and removed the em dashes and tidied up the surface-level tells.
Deloitte published a study finding that trust drops 144% if a user suspects that content has been created by AI. That is not a small number. And in a trust recession, where we are all flooded by content from every direction, trust is the currency that converts. If AI is quietly eroding that trust with your audience, the speed it saves you is not worth it.
Beyond the content quality issue, I started asking myself a harder question. Was a lot of what I was doing in AI just busy work? Was I getting a dopamine hit from putting things in and occasionally getting something back that looked reasonable, without it actually moving the needle in my business? It is the same mechanism as a slot machine. Random rewards. You never know when the lever will land on something good, and that unpredictability keeps you pulling.
The Nail in the Coffin
I want to be clear that no AI company is squeaky clean. If you are looking for a perfectly ethical tech company, you are not going to find one. But when news broke about ChatGPT’s contract with the US government, and the way that had been handled publicly versus privately, that was enough for me. A lot of people moved to Claude around the same time, and my reason was similar. Anthropic, the company behind Claude, has a philosopher on staff whose job is to think through what ethical AI actually looks like in practice. That is not nothing.
How I Migrated (and What I Learned)
A lot of the migration guides online will tell you to go to a certain tab and click export all. That did not work for me. I also quickly realised that I did not want to import everything I had ever put into ChatGPT into Claude indiscriminately. So I created a migration guide documenting exactly what I did and how I set up my Claude account from scratch. There is no opt-in required. Just DM me on Instagram and I will send it to you.
One thing worth knowing before you set up Claude is that it works on a token system. You have a certain amount of processing capacity available in any given period, and some account setups will burn through that before you have even asked a useful question. The guide covers this.
What I Actually Use Claude For
Podcast Production
This is where we use Claude the most, and it is a great example of how I think about AI in the business generally. I ideate and record every podcast episode entirely from my own brain. I set the topic, write a few notes, and record. My assistant Tanya then takes that transcript and uses a combination of Claude and Descript to produce the show notes and blog post. The original thought leadership, the language, the ideas, they all come from me. Claude handles the production layer, not the thinking layer.
Content and Copy Tweaking
I do not use Claude to create content. I use it to refine things I have already written. For example, I recently had a Meta ad that was attracting the wrong quality of leads. I took my existing caption to Claude and asked it to suggest where I could edit to better signal who the ad was targeting. It worked. That is the right use of the tool.
For sales pages, I will go through section by section and ask things like: can you tighten this headline grammatically without changing what I have written? Where could I create more open loops in this module description? How could I build more internal urgency in this section? I always specify that I do not want it rewritten wholesale. I want it to show me where to look and give me a suggestion, so I am still the one making the decision.
The Daily Briefing
I have set up a daily briefing in Claude that pulls together my daily, weekly, monthly, and quarterly task lists, my engagement hit list, client audit deadlines, upcoming birthdays for mastermind members, and my content calendar. It posts to a Notion page which then flows into our team Slack channel. It is not perfect yet, but it handles the mental load of keeping track of recurring tasks so my brain does not have to.
The Peace and Profit Magazine
Every month inside the Peace and Profit Mastermind we produce a magazine for our members. I upload transcripts from group calls and ask Claude to identify common themes and gaps in the curriculum. I have built a custom skill inside Claude to help with this process. The thinking is still mine. Claude surfaces patterns faster than I could manually.
Dashboards, Databases, and Grunt Work
My favourite thing to do in Claude is have it build Notion databases, HTML dashboards, and Kanban boards. If I have information sitting in a spreadsheet that needs to live somewhere more useful, Claude can move it there fast. I have also built educational games for my kids using Claude’s coding capability. The times tables game alone was worth the subscription.
What I Will Never Use Claude For
Content creation. Hooks. Email copy. Sales writing. Podcast ideas. Podcast scripts. Titles. Anything that requires my original voice, my thought leadership, my expertise, or my decision-making stays entirely with me.
I think about it this way. The goal of using Claude in my business is to replace the work a very capable intern would do: the formatting, the organising, the moving of information from one place to another. The goal is not to outsource the things that make my business mine.
If you want a business that runs simply, in part-time hours, without a large team and without you being the bottleneck, you can read more about how I approach that inside the Peace and Profit Mastermind. That model of running a lean, calm, and profitable business is exactly what shapes the way I use tools like Claude too.
The Bigger Question Worth Asking
There is a real risk that using large language models trains us out of thinking independently. I see it with my kids on school holidays. The more screen time they get, the harder it is for them to play creatively, to generate their own ideas, to be bored and work through it. I wonder whether the same thing happens to us as business owners when we outsource too much of our thinking to AI.
Sam Altman once suggested that humans would eventually pay for AI the way they pay for water because we will have trained our brains not to think independently. That concerns me. The originality, the nuance, the tiny specific details that make content convert, those are the things AI consistently misses. And they are also the things that are hardest to rebuild once you have stopped practising them.
AI scrapes what already exists on the internet and gives you back an average of it. If what already exists is largely mediocre, the output will be too. The question I keep coming back to is whether we are just adding to that pile.
Want the Migration Guide?
If you are thinking about moving from ChatGPT to Claude, or setting up Claude for the first time, DM me on Instagram at @robynbirkin and I will send you the guide. No opt-in. Just ask.