I’ve been writing Google Business Profile posts and social content for a small local business client for five years. I had a good working relationship, a steady retainer, and reasonable time commitment; it was the kind of freelance gig you don’t take for granted. Then I accepted a full-time position elsewhere, started exploring Claude’s newer features, and built a workflow so capable that my client asked me to hand it off as my replacement. Not to another freelancer but to the workflow itself. I was so good at automating processes that I actually automated the job entirely — and, I have complicated feelings about it.
My old workflow wasn’t “AI-written,” it was AI-assisted
Claude as a personal editor
Credit: Jonathon Jachura / MUO
For most of those five years, I wrote everything myself. Every GBP post, every caption, and every content calendar. I have a strong preference for human-written copy — not out of nostalgia, but because I think voice and specificity are hard to replicate, and both matter for local business content.
That said, I wasn’t working in a vacuum. As AI tools got better, I leaned heavily on Claude as a kind of always-available editor. I’d draft a batch of posts, then run them through Claude with the client’s style guide attached. It would flag tone inconsistencies, suggest sharper CTAs, and catch things that were technically fine but felt flat. When I was ideating topics for the month, Claude helped me brainstorm angles I wouldn’t have landed on myself. When I wasn’t sure if a post matched the client’s brand voice, I’d ask Claude to compare it against past examples.
The distinction felt important to me: I was writing, Claude was reviewing and optimizing. The content was mine. That balance worked well — good output, faster turnaround, and I never felt like I was handing the wheel over to something that didn’t understand the client. Any freelance writer who’s been on the fence about bringing AI into their process doesn’t have to go all-in. There’s a middle ground, and it’s a pretty comfortable place to work from.
Related
ChatGPT Is Getting Old: Try This Ultra-Smart AI Alternative Instead
If you’re bored with ChatGPT or want to try something new, Claude is the answer.
Then I went deep on Claude Skills and Cowork
Building a workflow I wasn’t expecting to give away
Taking the full-time job meant I had roughly a month to wrap up with my freelance client. We had a great five-year relationship, and I wanted to at least leave them with something useful. Around the same time I was working through the offboarding, I started digging into Claude Skills and Claude Cowork — Anthropic’s agentic desktop tool that brings Claude Code-style capabilities to people who have no interest in a command line. It runs locally in an isolated virtual machine, reads files from folders you designate, and handles multi-step tasks without you babysitting every move.
I put together a folder with the client’s brand guidelines, tone notes, a couple of dozen example posts I’d written over the years, and a basic prompt template. Then I pointed Cowork at it and asked it to generate a month of GBP posts. The first pass wasn’t perfect — it rarely is with any new workflow — but it was genuinely close. After a few rounds of prompt tweaking and updating the skill, the drafts had the right tone, the right topics, and the right length. They read like the content the client was used to getting from me.
That’s the moment I realized this wasn’t just a curiosity project anymore.
Developer
Anthropic PBC
Price model
Free, subscription available
I showed the client, and they asked me to finish what I started
What the automated GBP workflow actually looks like
I demoed the Cowork workflow on one of our offboarding calls, mostly to show them what was possible. Their reaction was pretty immediate: “If you’re already leaving, can you set this up so we can just run it ourselves?”
The workflow I handed off is simple enough that they don’t need a technical background to use it. There’s a folder on their machine with brand docs, style notes, a prompt template I wrote to create the Skill, and an archive of past posts. They open Cowork, load the folder, run the prompt, and get a draft content calendar back. Someone on their end — them, or a part-time hire — reviews it, makes edits, and schedules. The AI handles the first draft. A person closes the loop.
That review step was the part I pushed hardest on. AI tools are increasingly capable of stepping into creative and marketing roles, but nobody knows a business the way the people running it do. Skipping human review is where AI-generated content usually falls apart — not always because the writing is bad, but because it lacks the specific, ground-level knowledge that makes local content feel local.
I’ll also say this: for small businesses where budgets are tight, there’s a real argument for this kind of setup. Outsourcing content creation adds up fast. If a workflow like this can cut that overhead while keeping someone in the loop on quality, that’s a reasonable trade. I get why my client wanted it.
My take on AI-written content is pretty agnostic. For short-form posts where consistency and local visibility are the main goals, the stakes are low enough that a solid workflow plus human review covers you. It’s a different calculation than switching wholesale from tools you actually trust. For long-form work, I still write everything myself — there’s a texture and specificity to good long-form that I haven’t seen a model nail consistently.
When you automate well enough to make yourself redundant
There’s obvious irony in all of this, and I’m not going to pretend otherwise. Five years of building trust with a client, and my exit gift was a system designed to do without me. It’s a strange feeling.
What it made clear, though, is how much the barrier to this kind of automation has dropped. A year ago, a workflow like this would have taken real technical effort — custom integrations, developer time, or, at a minimum, a steep learning curve. Cowork puts it within reach of anyone willing to spend a few hours setting it up. For freelancers working in content or marketing, that shift is worth taking seriously. Not as a reason to panic, but as a reason to understand these tools before a client brings them up first.
