Email validation and compliance workflow

Every lifecycle email went through four to five rounds of review between the lifecycle team and Compliance before it could send. With no self-serve QA, each send became a back-and-forth that added days to production.

The original request was a spam checker, something to flag obvious deliverability risks before content reached Compliance. I built that first version. It worked technically, but the output was inconsistent.

The reason became clear once I put it next to the review itself: the workflow was checking spam signals, while Compliance was reviewing for specific regulatory language, prohibited claims, and jurisdiction-specific disclosure rules. I had designed it without the people who do that reviewing.

Stop iterating on the technical layer and redesign the workflow outward from the compliance review process, with Compliance as a design partner rather than a downstream approver.

Rebuilding from scratch rather than patching the existing workflow cost time upfront, and it was the only route to something the team would genuinely rely on.

Compliance review cycles per email: how many rounds of back-and-forth before sign-off. That was the friction the workflow existed to remove, and if the number didn’t drop it wasn’t working, whatever it produced technically.

I sat with Compliance before touching the workflow and documented how they actually reviewed content: what they looked for, what triggered a flag, what the rules were per content category, what varied by jurisdiction. That logic became the prompt architecture and validation layer.

It took four versions. V1 checked spam signals: functional, but aimed at the wrong question. V2 added compliance flag categories: better output, still inconsistent, because the model was interpreting rules rather than applying them. V3 brought Compliance in and replaced interpretation with explicit rule sets: reliable, but too verbose to use. V4 restructured the output so results were actionable in seconds.

Built in n8n so the lifecycle team could run validation themselves while drafting. Compliance reviews the output rather than the email, so the compliance question is answered before the email reaches them.