Writing it down doesn't mean it has left your head
Why most processes fail at automation: they were written down but never run. Do it, understand it, refine it, and only then automate the steps that earned it.

For years I had calendars, to-do lists, reminders and automations. Each held its piece just fine. But when several things ran at once (a handoff at work, an on-call incident, packing for a trip) I was still the one holding the connections between them.
Writing everything down hadn't taken it out of my head. It had only made a copy.
A list isn't a process
A list tells you what exists. A process knows which step you're on, what comes next, what depends on what, and what happened last time. The gap between the two is exactly the part that stays in someone's head. And it's the part that breaks the week that person is away.
The order that works
- Do it. Run the process for real, with steps on a screen instead of in memory.
- Understand it. See what actually happened: which steps were skipped, which took three times longer, where someone had to ask.
- Refine it. Keep what mattered, cut what didn't, and run it again.
- Then automate the steps that have become predictable. Only those.
Most automation projects start at the last step. They take a process that was never run with a shape, wire it to tools, and automate the chaos faster.
A guide, not an autopilot
With AI moving this fast, many tools try to put people on autopilot. I don't think that's the answer. The system should carry the order, the timing and the memory of last time. It should do what can be undone on its own, and ask before anything that can't.
That's how Routine Run works, and it's how we build for clients: first the process gets a form people can run, then AI takes the steps that earned it.
Which process in your business has been written down, but never actually run from the page?
Prema Vision designs and builds AI systems for companies, takes them to production, and keeps a person on them after launch.
More notes
Apple put a language model in every new iPhone. Here's what it's good for
What Apple's on-device model did well in Routine Run, where it failed quietly, and the test that keeps a prompt and its parser from drifting apart.
Let AI assistants into your product without handing them the keys
How Routine Run's MCP connector lets Claude read runs and act, while every change a model writes waits for a person to approve it on the phone.

