Practical AI marketing for marketers who want real systems, not theory. I run a 12-function marketing operation solo for under $300/month. Weekly: workflows, prompts, and tools you can implement.
One model switch cut my token costs in half.
|
Hey Reader, Quick question: when you open Claude, which model do you pick? Most marketers I talk to pick Sonnet or Opus by default and leave it there. A few go straight to Opus for everything because it feels like the "safe" choice. Both are costing them — in tokens, in speed, and in output quality on tasks the model wasn't built for. The thing is that picking the wrong tool is like using a forklift to move a coffee cup or hiring a lawyer to write a grocery list. I run a 12-function AI marketing operation — content, design, email, lead gen, analytics, the whole stack — and I test every tool I put in it. Over the past few months I ran the same task types across all four Claude models to see where each one actually performs. The results were clear enough that I built a 2x2 matrix around it. The two axes are effort and impact. Effort is how hard the model works before it answers. Think of it like asking a professor a question. Haiku gives you the first answer off the top of its head. Sonnet thinks for a minute. Opus works carefully through the problem. Fable explores multiple approaches, checks its reasoning, and comes back when it's sure. Impact is how much is riding on the output — and how much damage a wrong answer does. Map those two together and you get four clear zones: Low effort, low impact tasks — reformatting content, tagging leads, summarizing meeting notes — go to Haiku. It's fast and cheap and it doesn't need to think hard. Asking Opus to summarize a transcript is like sending a neurosurgeon to take out your recycling. High effort, low impact tasks — first draft copy, prompt workflow builds, competitor research — go to Sonnet. This is the everyday workhorse. Anthropic themselves recommend it as the default for most work, and after running it across dozens of workflows I agree. It's where you'll spend most of your time. Low effort, high impact work — marketing strategy, architecture decisions, system design, coding agents — goes to Opus. The stakes are high but the brief is focused. You don't need it to run a marathon; you need it to think clearly on something that matters. High effort, high impact work — complex multi-step automations, full agentic workflows, systems built to run without you — is where Fable earns its place. This is the model that plans, executes, and checks its own work across a long sequence of steps. It's not for everyday tasks. It's for the work that runs your operation while you're doing something else. For most marketing professionals, Sonnet with high effort turned on is the sweet spot. That combination covers roughly 70% of what I run through my own operation. The one thing to do this week: open Claude, look at what you ran through it in the last five days, and ask yourself honestly whether you matched the model to the task. If you've been defaulting to one model for everything, try routing one task type to Haiku — something high volume and low stakes — and compare the speed and cost. You'll feel the difference immediately. If you want to talk through how to set this up across your team's workflows specifically, you can grab a free call here. — Louis P.S. If you want the one-stop shop for mastering claude, I created this resource hub for you: www.masterclaude.ca |
Louis Cho
Practical AI marketing for marketers who want real systems, not theory. I run a 12-function marketing operation solo for under $300/month. Weekly: workflows, prompts, and tools you can implement.
Hi Reader, I'll say what I always say: Claude is the better tool for marketing. Better brand voice consistency. Better at following constraints. Better at holding a style guide across a long conversation. I use it every day. I recommend it to every client. But here's the reality: I talk to marketing leaders every week, and many are on ChatGPT. Budget decisions, IT approvals, enterprise contracts — the tool is already locked in. So I stopped arguing about which AI is best and started asking a...
Hey Reader, Most people think AI tools are neutral. Put a prompt in, get an answer out. The tool is just a tool. That's not how it works. Every major AI company was built by someone with a specific worldview — about what intelligence should do, what it should refuse, what it should prioritize when it has to choose. Those beliefs don't live in a mission statement. They get encoded into the training data, the guardrails, the reward functions, the things the model will and won't say. The output...
Hey Reader, Most people write prompts like wishes. "Write me a launch email." Then they judge Claude on whatever comes back, decide it's mediocre, and conclude AI isn't there yet. The model isn't the problem. You handed a contractor one sentence and expected a finished build. A working prompt is a brief, not a request. You'd never tell a freelancer "write a launch email" and walk away. You'd tell them who it's for, what it's selling, what you've already tried, what good looks like, and how...