AI for Automation
How to automate the repetitive work that currently takes human time. The tools, the workflow, and where to start.
What AI automation actually is
Most business automation is simple: if X happens, do Y. Zapier and Make have done this for years.
AI automation is different: if X happens, AI decides what Y should be, then does it. The AI makes a judgment call in the middle of the workflow.
This lets you automate things that previously required human judgment — categorizing support tickets, drafting personalized responses, summarizing documents, making routing decisions.
What you can automate today
**Customer support first drafts** — incoming message → AI drafts response → human reviews and sends
**Content repurposing** — new blog post → AI creates tweets, captions, and email → auto-schedule
**Research summarization** — incoming industry news → AI summarizes → delivered to inbox daily
**Lead qualification** — form submission → AI scores and categorizes → routes to right team member
**Inventory and ordering** — stock level drop → AI-generated purchase order → sent to supplier
The tools
n8n — Open source. Most powerful. Runs on your own server (free) or their cloud ($20/mo). Supports hundreds of integrations and has AI nodes built in.
Make (formerly Integromat) — Visual, no-code, cloud-based. Easier than n8n for beginners. Starts free.
Zapier AI — Easiest to start with. Less powerful. Good for simple workflows.
For most founders building their first automation: start with Make or Zapier. Graduate to n8n when you need more control.
How to build your first automation
Start with one repetitive manual task. Map every step:
1. What triggers it?
2. What information do you need?
3. What decision gets made?
4. What action is taken?
Then ask: which step(s) could AI handle? Build the simplest version first. Make it work. Then add complexity.
I run a [type of business]. I currently [describe the manual task] by hand. The steps are: 1. [step] 2. [step] 3. [step] Design an automation workflow that handles this. Specify: the trigger, which AI tool makes decisions at each step, which tools take actions, and how a human stays in the loop for any step that's high-stakes.
Human in the loop — always
Design every automation so you can override it. AI makes mistakes. Automation amplifies those mistakes at scale.
The pattern that works: AI does the work, human approves before it ships. As the AI proves reliable on a specific task, you can reduce oversight. But always keep the ability to intervene.
AI assists you at scale. It doesn't replace your judgment.
Start with one automation. Map the steps. Build the simplest version. Make it reliable. Then build the next one.