The best way to ship code? Frequently.
At least, that’s according to Conner Hiatt, director of engineering at Corporate Tools, a business operations tech company.
“We don’t make engineers wait around to ship good code,” Hiatt said. “When something is done, it should be releasable.”
Hiatt said what matters most is rarely having to roll back code that is shipped.
“If teams are shipping fast and we’re not rolling things back, that tells me the system is doing its job and the code quality is high,” Hiatt said.
Hiatt isn’t the only one who takes this approach. Research from Datadog found that teams that ship code in small, frequent batches tend to have less impact if something does go awry.
Built In spoke with Hiatt about how the team at Corporate Tools focuses on fast timelines, which empowers engineers to take ownership of their products.
Corporate Tools builds the behind-the-scenes tools that keep businesses running: software, LLC filings, registered agent service, website and domain registration, address services and more.
What’s your rule for releasing fast — and what KPI proves it works?
We don’t make engineers wait around to ship good code. When something is done, it should be releasable. The rule is pretty simple: Add the right integration-level test coverage, quality assurance test when the change warrants it and don’t turn deployment into a ceremony. Some teams ship right away. Some batch the week’s work and deploy together. Both are fine. The KPI is boring, which is the goal — we ship often and we almost never have to roll back.
Which standard or metric defines “quality” in your stack?
Quality is when engineers aren’t scared to ship. We use the normal stuff, GitLab pipelines, integration tests, production errors and QA when needed — but rollback rate is the one I like because it’s hard to fake. If teams are shipping fast and we’re not rolling things back, that tells me the system is doing its job and the code quality is high.
Name one recent AI/automation shipped and its impact on the team or business.
We built our own AI chat system in-house, running an open-source model on our own GPUs in our own data center. The important part is that it doesn’t just answer questions. It can actually do work through hundreds of MCP tools. Customers can buy a domain, set up hosting, fill out documents and move through real workflows inside chat. That took a team of engineers about four months and it turned AI from a fun side feature into something that can actually move the business and automate work for our customers.
