Nobody approves their own work. Not an agent either
Nobody approves their own work. Not an agent either.
An agent that carries out a task and then says itself that it is fine has looked at the work only once: with its own eyes. And what it missed when it made it, it also misses when it checks it.
That is not an AI problem. It is the reason an accountant checks someone else's books, a text gets a second reader, and a bridge is not signed off by the builder.
That is why every role in my company has two names: one who does the work and one who checks it. They are never the same. That is written on the role's card, in advance, and not only once something goes wrong.
It works in three places:
On the role itself. My LinkedIn agent writes the posts, but puts none of them live without my approval. The draft comes to me, I respond, it gets changed, and only then does it go into the queue. Until I say it is ready.
At the gate. Before work goes into the organisation or comes out of it, a separate part tests whether it meets the bar. Not the agent that made it, but something that does not know how hard it was to make, and so has no stake in making it look good.
At the exit. For everything that goes out, a check looks for anything that does not belong: no internal path, no amount that is not fixed, no commitment nobody made.
The result is that I can let an agent do a lot without watching every step. That sounds contradictory, but it is the reverse. Because there is a second pair of eyes, I do not have to be the first.
And it changes how I look at the one who does the work. I no longer ask "is it right?", because there is only one way to answer that. I ask: "what did you do, and where can the checker see it?"
A tidy way to start: for every task you hand off, write down two names, the one who does it and the one who checks it. If you can find only one, you have found a task without a check. That is the start of a list you can work through tomorrow.
Who checks the work of an agent or automation where you work, and is that someone who would also notice if something quietly goes wrong?
#AIagents #AI #Quality #Workflows
See also
I expect my agents to buy property and hire people within five years
I expect my agents to buy property, take over businesses and hire staff within five years, both agents and people. That is not a promise and not a plan. It is an expectation, and everything I build now is a step in that…
A file service of your own, built by an agent in nine minutes
Sharing files with people outside your organisation is often a hassle. Permissions in OneDrive or SharePoint, an account for someone who wants to download something just once, or an outside service where you do not know…
Two things you lean on: a green light and a report
When you let agents do the work, you get two kinds of signals back as a human. A test that shows green, and a report of what was done. You lean on both. And both can put your mind at ease without saying anything.
