Every time I come across messy operations, the cause is almost always the same: not a lack of effort, but processes that depend on someone's memory.
The symptoms are easy to spot
- Tasks get missed, and everyone notices only after the problem has grown
- Results vary depending on who happens to be doing the work
- Reports are always requested, never delivered on their own
- When the person with the best memory is off, everything slows down
None of this is about laziness. It's about systems that don't exist.
Why relying on memory always loses
Human memory is excellent for things that are important and rare. But operations consist of small, repetitive things — and that's exactly where memory leaks most. Not because people forget, but because attention gets divided.
The fix isn't "more discipline." The fix is moving the burden of remembering from people to systems.
The order of improvement I use
- Map it first. Write down the workflow as it actually is. Often the problem becomes obvious the moment it's written down.
- Pick one bottleneck. Don't fix everything at once. Take the one that breaks most often.
- Write a short SOP. One page. If it's longer, chances are nobody will read it.
- Add monitoring. So deviations surface early, not after they become problems.
- Measure, then expand. If one point works, add the next.
Where AI fits
AI doesn't replace operational understanding — AI automates the boring parts. Recording, summarizing, reminding, assembling reports.
What stays human: deciding what matters, and judging whether the results make sense. AI without operational direction just automates the mess.
That's why I build AI systems while still running the operations myself. So what gets automated is genuinely what's needed, not what merely sounds impressive.