A 20-minute pass over every automation you own, to find the ones running in the dark.
Companion to "The automation nobody sees." You don't need any tools open to start. A list and
twenty minutes.
Write down every automation you currently have running. Zaps, Make scenarios, n8n workflows, Kit
sequences, scheduled reports, dashboards, recurring exports, anything on a timer.
Include the ones you forgot about. Those are the point. If you built it more than three months
ago and haven't touched it since, it is a candidate.
Don't evaluate whether it works. Assume it works. Ask where the output lands.
| # | Question | The failure it catches |
|---|---|---|
| 1 | **Who receives the output?** | "The system" is not a person. If you can't name a human, it's running in the dark. |
| 2 | **Where does it arrive?** | An app you've stopped opening is the same as nowhere. |
| 3 | **When does it arrive, relative to when it's needed?** | A Monday report that matters on Friday is decoration. |
| 4 | **What does the receiver do next?** | If the answer is "notes it", there is no action and no value. |
| 5 | **What happens if it's ignored for a month?** | If nothing breaks and nobody notices, you have your answer. |
LANDS. A named person receives it, somewhere they look, in time to act. Leave it alone.
MISDELIVERED. The work is good, the destination is wrong. This is the valuable pile — the
fix is usually one field, not a rebuild. Change where it lands, not what it does.
DARK. Nobody receives it, or nobody acts on it, and nothing breaks when it stops.
Turn it off. Not "review later" — off. You can always turn it back on, and you won't.
Your own email list. A welcome sequence that fires perfectly into a promotions tab. A tag that
gets applied and never used for anything. A signup form wired to a lead magnet nobody has downloaded
this quarter.
Check the delivery, not the build. The build is almost always fine.
For each one, change exactly one thing: the destination. Not the logic, not the formatting, not the
schedule. Move it to where the person already is — the inbox they read, the app they open, the
channel they answer — and change nothing else.
Then leave it a fortnight and check whether anyone acted. Delivery is testable. Test it.
If this turns something up — especially something that had been firing into the dark for months —
reply and tell me. I'm collecting examples.
If you'd rather I just did it
I run this exact pass as a fixed-fee job. You send me a list of what you have running, I come back with what has silently stopped landing, with evidence for each one, and what to do about it.
I ran it on my own setup last month and it found seven things quietly failing. One of them was a newsletter that had gone out 228 times with no clickable link in it.
I'm taking one of these to start, so I can do it properly rather than at scale. If you want it to be yours, reply to this email and tell me roughly how many automations you're running. I'll come back with a price.
No call. It's all async.
Automation Wizard/Soloneer - Exploring how we can use the latest technologies such as AI, automation tools like n8n, and 3D printing to make our lives easier and better.
The hardest part of building an AI agent system isn't the technology. It's the blank page. You know what you want: an agent that knows its role, stays in its lane, handles recurring work without you micromanaging it. But when you sit down to write the system prompt, you stare at the cursor. What do you actually put in there? Over the past several months I've built a 5-agent council that runs my YouTube production pipeline. There's a CEO agent that sets direction and delegates. A technical...
Ten weeks since my last email. The automation guy went quiet again, and this time the reason is a bit embarrassing. I built the AI Agent Workflow Starter Kit and wrote this email announcing it. Then I never pressed send. It has been sitting in drafts for months while I got on with everything else. Turns out being busy is a very reliable way to ship nothing. So here it is, as I wrote it. The hardest part of building an AI agent system isn't the technology. It's the blank page. You know what...
I once spent £300 on an Iron Man figure because I "earned it." This was my productivity reward system. Hit a revenue milestone, buy a collectible. Complete a major project, add to the collection. Sounds smart, right? Gamify your work. Create incentives. Make the grind feel like a game. Here's what actually happened: Month 1: Hit the milestone, bought the figure, felt genuinely motivated. Month 2: Missed the milestone, bought the figure anyway "for future motivation." Month 3: Tight month...