Every SMB running custom software systems has a list somewhere of things they “should really” do to their software. It’s usually buried in a backlog tool, a spreadsheet, or a half-remembered conversation. Customer self-service tweaks. Internal dashboards that would save hours every week. Cleaning up reporting. Automating the thing someone still does manually every Friday afternoon. None of it is controversial. All of it lost, over and over, to whatever was on fire that week or to the reality that it would take a senior developer a couple of months to get it done.

Thanks to AI, that math is changing.

Over the last year at Trailhead, working with companies our own size, we’ve watched that same pile or work start to move for the first time in years. Work that used to require careful scoping, multiple handoffs, and a meaningful budget now gets done in a couple of weeks or a few focused days when someone knows how to aim AI at it. Not everything, and not always cleanly, but enough that the backlog is no longer theoretical.

This session is a working playbook for clearing that list. I’ll walk through real examples like generating admin tools on top of existing databases, building one-off integrations between systems that never quite justified a project, tightening up customer onboarding flows, and turning tribal knowledge into usable internal tools. You’ll leave with a prioritized checklist of high-value backlog items, realistic time and cost ranges for each, and a simple decision framework for what’s worth doing now versus leaving on the shelf.

If you’ve ever said “we should really get to that someday,” this is how you get to zero.