Automation saves time on the task and not necessarily on the day. My workflows ran 10,070 times in five days and failed 8 of those times, so almost nothing breaks. About two alert emails still land in my inbox daily, at times I don’t pick, and each one needs a decision from me. Automation handed me capacity instead of free hours, and I spent every bit of it starting more projects.
The morning I planned and did not get
I wake up knowing what I want to work on. A new app, a post, a workflow I’ve been thinking about in the shower. Then I open my email and there are failures sitting there. A renderer refused an image. A sync timed out. Something I built in July wants attention before anything else happens.
So I work the queue instead. And if Claude Code is having an off day, which happens, a twenty minute fix turns into two hours of me going “no, not like that” into a terminal. By the time I look up, the thing I actually wanted to build is still sitting there untouched, and I’m annoyed.
I built all of this. Every single piece. Which makes the frustration extra funny, and also extra hard to blame on anyone.
Does automation save time?
Automation saves time on the task and costs time on the system. The individual job gets faster, permanently. The pile of jobs gets bigger, also permanently, because a thing running well is an invitation to build the next one.
My setup right now looks like this, and every number is pulled from the live box, not estimated:
- 221 workflows built, 138 of them active
- First one built on December 4, 2025
- 22 new workflows in September alone, through the 16th
- 10,070 executions in five days, roughly 1,900 runs a day
- 8 failures in those five days, across 5 workflows
Look at that September number. Nine months in, with 138 things already running, and I added 22 more in sixteen days. I did that to myself.
Why do 8 failures still eat a whole morning?
My failure rate is 0.08%. Out of every 10,000 runs, about 8 go wrong. By any normal measure my stuff is boringly reliable.
And I still get roughly two alert emails a day. Over the last month, about 60 of them. The number of failures is tiny. The number of interruptions is not, because failures don’t arrive in a batch at a time I chose. They arrive whenever, and every one needs a human to look at it and decide whether it matters.
That’s the part I got wrong when I started. I thought the risk was fragility, that things would break constantly and I’d be patching all day. They don’t break. They just have standing permission to interrupt me, and I’m the one who granted it when I pointed all the alerts at my inbox.
Same handful of workflows send most of the mail, too. My image renderer refuses a prompt, retries, refuses again, and emails me about it. Fixing those few would cut the noise a lot. I know this. I have known it for weeks. Ask me how far up the list it is.
Automating for a business is a different sport
Somewhere in the middle of complaining about this out loud, I said something I keep coming back to. Automating for your business is a different activity from automating because you like building things. Same tools, same skills, completely different finish line.
| Automating for the business | Automating because you like building | |
|---|---|---|
| Why you start | A task keeps eating your week | You want to see if you can |
| When you stop | When the task runs without you | You don’t, the next idea is already open |
| What success looks like | Fewer hours on the same work | More things existing |
| The failure mode | You skip automating and stay stuck | The pile grows faster than you maintain it |
| What to measure | Hours the task used to take | How many alerts you get a week |
I’m firmly in the right-hand column, and I’m not pretending otherwise. If you’re in the left one, the ROI check is the whole job: build the thing, confirm the task runs without you, walk away. Being done is allowed.
If you’re in the right one like me, the ROI math doesn’t help, because payback was never the point. The thing worth watching is your alert volume. Hours saved is the wrong number for a builder. Interruptions per week is the right one.
The layer I built without noticing
I’ve spent months building things whose only job is to carry the weight of the other things.
An error workflow so a failure emails me instead of sitting silent for a week. A schedule watchdog after I found a workflow that had never run once in seven weeks. A to-do list for Claude Code so work happens before I sit down. A brain dump tool so the ideas stop rattling around in my head at 11pm.
Every one of those helps. And every one of those is also another thing I maintain, with its own way of failing, and in the error workflow’s case, its own contribution to my inbox. The alert that ruins my morning is a tool I built, working exactly as designed, doing the job I asked for.
I don’t regret any of it. I’d rather get the email than find out in October. It’s still funny in a bleak way that my solution to too many moving parts was more moving parts.
What I still haven’t solved
I like this. I’ve been a workaholic my whole adult life, so a setup where I can build whatever I want, whenever I want, and watch it run is close to a dream for me. I’m not writing a cautionary tale and I’m not quitting.
The overwhelm is real anyway. Some days it’s a lot, and it doesn’t feel good, and I don’t have a clean fix to hand you. The honest state of things is that I’m still figuring out how to keep the building and lose the ambush.
That shape isn’t specific to workflows. It’s the same reason you can resent the upkeep on something you genuinely wanted: the wanting was one decision and the upkeep is every day.
The one change I’m sure about: stop sending every alert to the inbox I work out of. Route the noisy ones somewhere I check on purpose, once, at a time I pick. My morning should belong to me and not to a renderer having a bad night in July.
Frequently asked questions
Does automation save time?
Yes, on the task, and no, on your day. The job itself gets permanently faster. Whether your week gets lighter depends entirely on what you do with the space, and if you’re a builder you’ll fill it with more building.
How much maintenance do automations need?
Less repair than you’d think, more attention than you’d like. My 138 active workflows failed 8 times in five days, which is nothing, and still generated about 60 alert emails last month. The upkeep cost shows up as interruptions, not as broken things.
Why do I feel busier after automating?
Because automation raises your ceiling instead of lowering your workload. You finish the same work sooner, then start something new with the room you freed up, and the new thing needs maintaining too. You end up with more room and the same amount of tired.
How many automations is too many?
Watch your alert volume, not your workflow count. If the notifications from your existing setup regularly take over the day you planned, you’ve hit your limit, whether you have 12 workflows or 138. The number is personal and it has nothing to do with how many you can build.
Should I automate if I love building things?
Absolutely, just go in knowing you’ll build more, not less. Set the stopping rule up front, decide where alerts land before you turn anything on, and check your interruption count every so often. Building is the fun part, and you can keep it fun by not letting it interrupt itself.
Try it yourself
Go look at where your automation alerts land. If they arrive in the inbox you work from, your systems can take your morning any time they want, and a 0.08% failure rate will still feel like chaos. Move the noisy ones to a channel you open on purpose. The failures aren’t the problem, the ambush is.
Subscribe if you haven’t already, and I’ll keep writing these up as I go, including the parts where I’m still stuck.
A tool I built
Side Hustle Garden
Log every side hustle payment and watch it knock out your monthly bills, smallest first.
Plant your first dollar →I run an affiliate program on the tools I build. Approved affiliates earn 20% on what their referrals pay, for up to a year. US only for now. See the program →
