The Hard Part of Automation Is Seeing It
I haven’t written much this month. I’ve been busy, although to be fair, quite a lot of that “busy” was running.
I’m properly back at it now, and the injury finally seems to be more or less under control. Touch wood.
The other thing taking up some of my time has been talking to people about automating the boring parts of their work through skills and agents. And I keep running (pun intended) into the same problem.
Explaining the technology is usually the easy part. Helping someone see what it could actually do for them is much harder.
You can explain that an agent can read something, make a decision, take action, check the result, and keep going until the job is done. People understand that. They nod. It sounds useful.
Then you ask what they would actually hand over from their own week, and they go blank.
The gap between “this is powerful” and “this annoying thing I do every Tuesday could simply stop being my problem” is bigger than you would think.
Which makes sense. You need some idea of what agents are capable of, but you also need to look at your own work differently. You need to notice that the thing you do every week is not just “work”. It is a process. It has steps. It has inputs and outputs. And maybe you do not need to be the person doing all of it.
Most people have never been taught to look at their work that way. Nobody has shown them what a good automation opportunity looks like, so of course they are not spotting them everywhere.
So I built something to hopefully ease that process a bit. It's rough, and definitely not perfect, but is a small catalogue of skill and automation ideas:
You can browse through examples of the kind of boring work people are starting to hand off. Hopefully one of them makes you think, “Wait, I do something like that too.”
The bit I care about most is the prompt.
There is a copy-paste prompt on the site that turns the agent into a brainstorming partner. It asks about your work, the things that take too much time, and the stuff you keep putting off because it is dull or repetitive. Then it helps you figure out which parts might actually be worth automating.
Normally, that is the conversation I would have with someone across a table, trying to tease out the useful ideas.
Now the agent can do it.
Which was sort of the point anyway.
Have a play with it. And if you end up automating something useful, send it my way. I collect these.
Cheers.