Pick what to automate first, then pick the tool. Most teams reverse that and pay for a platform that encodes chaos.
List three candidate loops. Score them. Automate the winner’s capture path, not all three.
Choose the platform after the winner is written down. Feature lists are how you buy the wrong seat.
flowchart LR
C["Candidates"] --> S["Score"]
S --> W["One winner"]
W --> T["Tool"]
T --> B["Build capture only"]

Score 1 (painful but rare) to 5 (daily, rule-based, currently copy-paste). Highest total goes first. Ties break toward the loop that touches money or a client.
| Candidate | Volume | Rule-clarity | Time saved | Risk if wrong | Total |
|---|---|---|---|---|---|
| Inbound lead intake | |||||
| Publish / ship checklist | |||||
| Invoice or payment chase |
New /contact row to a Sheet, ping Simeon, same-day human reply.
n8n, Zapier, and Make can all POST a form to a sheet. The difference is where the work lives, how you handle secrets, and whether you will outgrow the task model. Deep comparison is in the guide. This table is the pick.
| If this is true | Start here | Do not start here |
|---|---|---|
| You want the workflow as a file you can version, self-host, or inspect | n8n | A black-box zap you cannot diff |
| Non-technical owners must edit mapping in a vendor UI this week | Zapier or Make | A half-finished n8n instance nobody will open |
| The step needs a large language model (LLM) draft | Any of the three, plus a human-in-the-loop (HITL) pause | Silent send from the model |
| Secrets, production data, or a kill switch matter | n8n on infra you control, or a vendor with env isolation you understand | Keys pasted into a node and screenshotted into Slack |