I spent over eight years at Amazon Web Services, and the useful part wasn't the title. It was that the work kept handing me the same problems a small law firm hits every week: work stacking up in the wrong spots, people doing tedious things a computer should handle, two systems that won't talk to each other.
What surprised me is how little the fix changes with size. Getting waste out of fifty warehouses and getting it out of a ten-person practice come down to the same few moves. You just aim them at different problems.
So here's what carried over.
From warehouses to workflows
When I did capacity planning for Amazon Fresh, a lot of the job was hunting down places where people's time was leaking out. A paralegal stitching three spreadsheets together by hand is the same problem as a warehouse lead tracking inventory on a clipboard. Hours going in, nothing valuable coming out.
So I'd build reporting that did all of that automatically. Not to get rid of anyone, but to hand people back the hours for the work that actually needed a human. And the payoff was predictable: measure what's being done by hand, automate the repetitive part, then check the clock again.
In a firm that usually comes down to a handful of things. Intake that files itself instead of someone retyping details out of emails and voicemails into your case system. New files that route to the right person the second they land, with nobody sitting in the middle playing traffic cop. Compliance checks that live inside the workflow and catch problems early, so you're not booking a separate audit to go find them.
Same move every time. Find the by-hand work, automate it, then confirm it actually gave the time back.
Building things that hold up when it counts
A big chunk of my career was spent in the moments where something breaks in production, an executive wants an answer in the next ten minutes, and the meter is running the whole time. You learn fast that reliability is basically the entire job.
That's where I think most small-firm automation quietly falls apart. Someone wires up a Zapier flow for intake, it works nine times out of ten, everybody's happy. Then the tenth time a lead never gets logged, or a document just vanishes, or an email routes into nowhere, and nobody catches it for a week.
When I build something for a firm it has to work every time, and not because your stakes look like a warehouse's. It's because the person on the other end feels the failure just as directly. Your paralegal shouldn't be babysitting a workflow, half-wondering if something slipped. You shouldn't be up at night wondering whether that statute of limitations reminder actually went out. Building things people can lean on like that came straight out of running systems where "it usually works" didn't cut it.
Deciding what's actually worth automating
This one maps over almost perfectly. When you're running a complicated operation you're always short on something. At Amazon it was budget. At your firm it's people. You can't just hire until the problem disappears, so you're forced to choose: what's genuinely worth automating, and where does it buy you the most breathing room?
I made those calls at scale for years. Which processes are really driving value, which ones are stable enough to hand to a machine, where you keep a person in the loop no matter what. It's the same conversation I end up having with every firm. Intake? Almost always worth it. The actual legal thinking? No, and I'll talk you out of it if you push. Deadline tracking is the one I lean on hardest, because blowing a statute of limitations is both catastrophic and completely avoidable. Having made those trade-offs at scale is a lot of why I'm useful to a small practice. I've already watched which automation pays for itself and which stuff sounds sharp in a pitch and then creates more cleanup than it ever saved.
What that means for your firm
I built OpExCell around a promise that's a little uncomfortable for me: no results, no pay. My income rides on whether your workflows actually improve and your time actually comes back. I'm only willing to work that way because I've lived with the fallout of operational calls at scale, so I know what a real result looks like and how to prove one.
The work usually ends up being some mix of intake that captures leads and drops them into your case system on its own, document handling that sorts and routes files and flags anything with a compliance risk, email triage that drafts replies to the routine stuff, deadline tracking you don't have to think about, and billing that pulls time entries straight out of the work instead of making someone key them in later.
The exact build shifts from firm to firm. What doesn't shift is the approach: find the repetitive work, build something you can trust to handle it, then actually measure whether it gave your people their time back.
If you've got ten people or fewer and the admin work is slowly eating the practice, that's the conversation I want to have.