The system I built is being retired. That is the success story.

In 2016, a year into my job as a physics educator at the Schwartz/Reisman Science Education Center, I took on a side role: managing equipment orders for lab experiments.

The process: teachers opened a catalog file, picked their experiments, and emailed me the details by Tuesday, a week before they needed the equipment. I copied every email into a spreadsheet by hand. On Wednesdays I sat with the lab technicians to check what could actually be supplied. Then I emailed every teacher a confirmation.

Two months in, I'd had enough. Not of the work, of the waste. The whole loop was a machine pretending to be a person.

I'd rather spend 100 hours automating a task than keep doing it by hand for an hour a week, forever. So I brainstormed with teachers, sketched solutions, and caught the center's director in a hallway. He bought the idea, with a condition that scared me: if we were doing this, the system would also cover budgets, procurement, and electronic manager sign-off. A full operations platform. I had signed up for equipment orders.

I said yes anyway.

Over the next year I interviewed people at every level of the organization, mapped needs, and wrote a spec. With a consultant the director paired me with, we shortlisted vendors and collected quotes. The two enterprise heavyweights could do everything except the one thing we needed, lab-equipment ordering, and explained politely that we would adapt our workflows to theirs. A small local vendor had already built equipment ordering for university physics labs, offered to adapt it to us, and quoted a fraction of the price. Easy choice.

Four months later the system was live with a complete database. We ran the organization's first-ever inventory count on it. Over the summer I trained everyone: teachers, lab techs, admin staff, finance, IT. In September 2017 we switched, and the first improvement shipped almost immediately: a simple order form that replaced the emails to me entirely.

What happened next is the part I keep retelling.

The doubters, mostly people tired of learning yet another system, each got a short 1:1 walkthrough where they saw what changed for them: order three days ahead instead of a week, see the full catalog, get a warning when equipment runs short. The resistance dissolved in weeks. We started with 15 teachers and about 940 orders a year. By last year my center alone had 38 teachers placing about 8,000 orders a year. And as the network grew from 2 centers to more than 10, every new center chose the system too. My part in that spread was small and deliberate: train one lead teacher per center, give them real ownership, and step away. The patterns spread without me in the room.

My colleagues never stopped joking about the punchline: I automated my own role out of existence. The software worked, the training held, and the job I had been given had almost nothing left in it.

Not everything worked. A barcode-based fault-tracking module went unused. A substitute-equipment mechanism was adopted only in part. And the budget module, the director's original condition, was eventually abandoned: I didn't know the finance domain well enough to spec it properly, and the person who depended on it never fully trusted it. That failure taught me to filter requirements as hard as I filter ideas, and when a requirement sits outside my domain, to either learn the domain properly or push back early.

Now, after eight years in production, the system is being retired. A teacher at another center, a former software team lead, built a successor with a cleaner interface and integrations we could never have supported. Feature for feature, it is almost exactly what we built in 2017. The tool is dying. The ways of working it created are the network's standard.

That, to me, is what adoption means. Not the launch and not the demo, but what people are still doing years after the tool that taught them is gone.