Automating a spec-to-3D-print workflow

Guerraco Ballistics is an owner-operator-led specialty manufacturer. Jorge Guerra runs a tight team that ships physical product to a discerning market. Like most businesses at that size, the time-sink isn't the work itself; it's the recurring rituals around the work.

One specific recurring workflow, getting from spec to 3D-printed part, was eating real time. The kind of weekly chore that takes hours, has to be right, and lived only in one person's head. It was the sort of work AI was supposed to make easy, but the off-the-shelf tools didn't fit the actual shape of the workflow.

A working pattern, not a tool. We took the existing process Jorge and the team already trusted, identified the parts that were repeatable, and built a synthetic-team workflow around them: humans staying in charge of judgment and quality, AI handling the rule-based work that didn't need a human brain in the loop.

The capability was documented in plain English. The team got a walkthrough. Where the process needed to flex, we wrote down how to flex it.

The weekly bottleneck stopped being a bottleneck. The capability outlived the engagement; it's still in use, owned by the team, extended by the team.

We're early enough in this engagement that named outcome numbers (hours saved, error rate, throughput) are still being measured. We'll publish them when we have a clean read, not before.

Guerraco has more than one bottleneck. We solved one. The next ones (broader inventory work, customer-comms automation, demand forecasting) are still entirely human-driven. Whether we tackle any of them next is a decision for Jorge to make, not us.

That's the model. We don't expand scope to fill time. We solve a thing, hand it over, and let the next conversation start when the business is ready.

Three reasons we believe the work stuck:

• Jorge stayed in the loop on every decision. The capability reflects how he wants the work done, not how an outside consultant would have done it.
• The deliverable was the working pattern, not a tool license. There's no vendor lock-in and no ongoing fee to keep using what we built.
• Documentation was written for the team that runs the business, not for us to refer back to.