Founder notes
Who checks the automation?
By Djordje Milićević · Thinking Mans Systems
I’ve had over 50 interviews since May 2026 with business people building and buying AI.
One question that kept resurfacing was: who checks its work?
But also, who manages it, who actually wants it, and does it make the problem disappear or just move it somewhere else?
A few convos and lessons I remember in particular:
1. Understand the business before talking about the technology
I spoke to someone who spent around 30 years in UK hospitality.
He said operators are already under pressure from staffing costs, taxes, margins, inflation, all of it.
So when you approach them, they need to feel pretty quickly that you understand the reality they’re dealing with.
His advice was basically: understand the finances, the costs, the problems the industry is facing, and show that from the start.
He literally said:
“Empathize in some way.”
A lot of them are skeptical of this stuff anyway.
Coming in and explaining how good your automation is probably isn’t going to do much if it sounds like you don’t understand their business.
2. Automation can just move the work somewhere else and/or have unexpected consequences
A fractional product consultant gave me a good example.
Imagine a nurse who’s already stretched thin, running all over the place at work.
Someone automates part of her job, but now she has to check everything the system produces.
She wasn’t trained for that, and she doesn’t have enough time already.
He called this a performative gesture: we automate the process, tell the humans to check the output, and call it a day.
And there’s another problem he raised that most people aren’t thinking about.
The concern is that the longer people spend checking instead of actually doing the work, the more their own knowledge can deteriorate.
Eventually, who’s qualified to catch the mistake... no one?
3. The cost of managing the automation matters, a lot
A systems advisor who works with founders asks three questions before saying yes to any automation:
- Is this actually going to reduce my load, or is it one more thing to manage?
- How much will it demand from me to make it real?
- Will it be as easy for me as it is for the person demoing it?
Much better questions for a business than just:
“Can this be automated?”
4. Sometimes the boring solution is the better one
In one example shared with me, an insurance agency had Post-it notes on a wall.
Someone put a camera above them that scans the notes, emails the owner for confirmation and updates the calendar. That’s basically it.
Then the owner gets one email, checks it and says yes.
The client was happy because the problem disappeared.
No particularly fancy tech involved, but good results still.
So, before you sign off on any automation, ask who checks the output and how long that takes them.
(Besides like 70 other questions, but that’s a story for another post.)