Most of what we know about building AI automations used to live in scoping calls and Slack threads. This week we wrote it down and published it.
What shipped
- Build guides. One page per workflow we get asked about, from invoice processing to tenant maintenance requests. Each one covers the architecture, the stack, what breaks in production, and what it costs.
- Published pricing. Price ranges for every kind of project we take on, with the factors that move a project up or down the range. See what AI projects cost.
- Free n8n templates. Fifty importable workflows that pair with the guides. They use Claude for the AI step, validate results in code, and send anything uncertain to a person. Browse the templates.
- Tool notes and comparisons. Integration notes for the business tools our clients use, and honest comparisons such as n8n vs Zapier.
- A five-minute readiness assessment that runs in the browser and recommends a first project. Take it here.
Why
More of our clients now start with a question to ChatGPT or Claude rather than a Google search. Those assistants answer with whatever specific, current, well-structured material they can find. We would rather they find ours.
So every page also has a plain-markdown version at the same address ending in .md, and the site publishes an index for AI assistants at /llms.txt.
What we learned
Writing down the failure modes was the most useful part. Every guide has a “what breaks in production” section, and most of those bullets came from things that actually broke on client projects: duplicate invoices, vendor names that drift, webhooks that fire twice.
Next week: the first round of fixes from readers, and more templates.