Is this worth building?
Outcome
You can tell which tasks are worth the effort of building a reliable AI workflow, and which aren't.
Concept
Once you know how to build something reliable, the next question is which tasks deserve that effort — building to 95 percent is real work, and you don't do it for everything. The tasks worth it tend to share four traits: they happen often, they're slow or annoying by hand, the information you need is already available, and you can actually tell whether the output is right. A task that's rare, or that you can't check, usually isn't worth automating.
Weak approach vs. strong approach
The ask
I want to automate the way I summarize competitor news. It would be cool to have Claude just do that for me.
What comes back
Sure, that sounds useful — but "cool to automate" isn't a strong enough reason to invest the work. Competitor news varies wildly in format and source, you probably don't check it on a fixed schedule, and it's genuinely hard to tell whether a summary missed something important. You could build this and never quite trust it, which means you'd still skim the originals anyway. The effort of making it reliable might outstrip the time it saves.
Try it
List a few tasks you do regularly. Score each on four things: how often, how painful, whether the information is already available, and whether you can check the result. The one that scores high on all four is your candidate.
Takeaway
Reliable AI work is worth building only for tasks that are frequent, painful, data-ready, and checkable. Choosing the right one is half the skill.
Your turn
What are you thinking of building?
Or start from one of these
How often do you do this?
How slow or annoying is it by hand?
Is the information you'd need already available?
Can you tell at a glance if the output is right?
Score all four to make the call.
4 more to go — the dial settles once every factor is set.