Section 1 — Hook + Ethics framing (~430 words) Most automation projects in regulated industries hit the same wall eventually: a CAPTCHA on an internal portal blocks the very automation the team is trying to build. In our case, the ops team needed to interact with one of our company's internal portals dozens of times per day. The portal — built by an internal team, used only by employees — gates access with a 6-digit numeric CAPTCHA on every login. Reasonable security choice for the original threat model. Not so reasonable for the team that needs to script repetitive workflows on top of it. The right fix would have been to add a service-account API to the portal team's backlog. The realistic fix, given the timeline, was to teach a small ML model to read those CAPTCHAs reliably so our automation script could move past the login screen.…