The hardest part of automating work isn't the building. It's choosing what to build. Pick the wrong task and you spend real effort solving a problem that wasn't costing you much, or — worse — automating away a piece of judgment that should have stayed human. Pick the right one and it pays back quietly, week after week, for years.
Most people pick by pain: whatever's most annoying right now. That instinct is usually wrong. Here's the test we actually use.
The three things that make a task a good candidate
A task is worth automating when all three of these are true:
- It's frequent. The payback compounds with repetition. Something that happens every week, or every day, can justify the build many times over. A once-a-quarter task almost never does, however much it grates.
- The routine and the judgment are separable. The best candidates have a thick layer of rote work — gathering, formatting, cross-checking, drafting — wrapped around a thin core of real decision. You automate the wrapper and leave the core to a person.
- A person can still own the outcome. If the task can keep a human reviewing and deciding before anything takes effect, it's safe to build. If it would require the machine to make the final call on something costly, it's not a first candidate — maybe not a candidate at all.
Frequency makes it worth it. A separable routine makes it buildable. A human still owning the outcome makes it safe. Miss any one and reconsider.
Why "most painful" is the wrong filter
The task people complain about loudest is often painful precisely because it's all judgment — the tricky negotiation, the hard diagnostic call, the high-stakes decision. That pain is real, but it's not the kind automation should touch. Pointing a tool at it either fails or quietly replaces judgment you needed.
The work actually worth automating is usually the opposite: not dramatic, just relentless. The background reconstruction, the re-formatting, the same draft from scratch again. Nobody puts it on a list because it doesn't feel like a crisis — it just eats hours.
Do the back-of-envelope before you build
Before committing to anything, estimate it honestly:
- How many hours a week does the routine part really take? (Not the whole task — the part you'd actually hand off.)
- How often does it recur, and across how many people?
- Who's doing it — and what could they do with that time instead?
A task that pulls a few hours a week from an expensive, hard-to-replace person pays back fast. A rare task, or one that only saves minutes, rarely justifies a custom build no matter how irritating it is. Being willing to say "this one isn't worth it" is what keeps the ones you do build honest.
The candidate you'll miss
One pattern worth naming: the best candidate is often invisible. It's not the task people complain about — it's the quiet, constant, low-status work that's so woven into the day nobody thinks to mention it. Finding it usually takes a short, honest look at where the hours actually go, rather than a list of complaints.
That's the conversation we start with: not "what do you want to automate," but "where is your time actually going." The answer is frequently a surprise.
Where Honewright fits
Finding the right work to take off someone's plate is the first thing we do — before any building. We look for the routine that's eating skilled people's time, the part that can be handled while a person stays in charge of the decisions. Sometimes the honest answer is that nothing's worth building yet, and we'll tell you that too.
If you're not sure where to start, that's the normal place to start. Tell us what's eating your time and we'll help you find the candidate worth the effort.