AI WORLD HQ
It's Monday morning, you've got 80 unread emails and a spreadsheet due by lunch — and the automation you'd normally set up now wants you to pick which AI "brain" should read, summarize, and write for you. The interesting question is which brain fits which kind of work.
Why there are so many models to choose from
Every major AI company builds its own LLM, which is short for "large language model" — the trained brain that powers chatbots like ChatGPT. OpenAI has its GPT family, Google has Gemini, Anthropic has Claude, and open-source efforts like Mistral are out there too. Each was trained with slightly different priorities, which is why no single "best" answer exists.
A useful mental model: think of them like cars. They all drive, they all have wheels, but a pickup, a sedan, and a minivan each shine in different situations. The same is roughly true of AI models — they overlap a lot, but they each have a few areas where they tend to perform well.
How automation platforms fit into this
Tools like Zapier, Make, and n8n are essentially switchboards. They connect Gmail, Slack, Google Sheets, Notion, your CRM, and hundreds of other apps, and trigger actions when something happens. When you add an "AI" step to a workflow, the platform asks which model should do the thinking — usually through a simple dropdown. You don't need to write code, you don't need to manage API keys (an API is just a structured way for two programs to exchange information — the plumbing that happens behind the scenes).
What each model family tends to be good at
This is a rough guide, not a ranking:
- OpenAI's GPT family — widely adopted, with the most documented use cases and the longest track record. A safe default if you don't have specific needs.
- Google's Gemini family — built strong for images, video, and audio from the start. Useful when your automation needs to "look" at a screenshot, a chart, or a short video clip.
- Anthropic's Claude family — designed for very long documents and careful step-by-step reasoning. A common pick for 200-page PDFs, dense contracts, or multi-step analysis.
- Open-source options like Mistral — popular with teams who want more control over where their data lives or who need to run the model on their own servers.
On most everyday jobs — drafting a reply, summarizing an email, sorting a row in a spreadsheet — they all do the job fine. Your choice usually only matters once you bump into the edges: a giant attachment, an unusual file type, a privacy rule.
A simple way to start today
Pick one task you do every week — sorting customer emails, drafting meeting notes, pulling action items from a transcript. Open your automation tool, choose a model in the AI step dropdown, and ask it to do that single thing. Use it for a week. If it saves you time, expand it to the next task. If it stumbles, try a different model in the same step. The fastest way to find your fit is to run it on your real work, not to read another comparison.
That is the whole approach: understand the lay of the land, start small, and let your actual results guide the choice.