A first-pass set of project plans for the ideas in Redline's list — what each one really is, how we'd build it, what it depends on, and the honest cost to own it. AI-assisted development makes the building fast; these specs focus on doing it right, not on hype.
Concrete, measurable, fast to ship. Each replaces clear manual labor or rides a pattern we already run.
A vision model screens every first photo for framing on upload, auto-emails the MS + RM on clear fails, and routes only the borderline few percent to a human. Replaces a daily 2,000+ photo manual review.
A scheduled job pulls billing from Predian and hours from Rippling and sends each Merchandising Specialist a personalized Slack recap — billing, VINs, utilization, goal attainment. Reuses our existing cron→Slack pattern.
Drafts the 2–3 weekly coverage updates RMs write by hand (15–30 min each) — live numbers, today's hitlist, weather-aware language — and posts to Slack with one-click approve. RM stays in the loop.
Real value, real engineering. These are projects, not weekend scripts — and the honest cost lives in ownership, not the first build.
A genuine vehicle-routing / territory-balancing problem (OR-Tools), not a chatbot. Balances 100+ routes by demand, home address, distance and utilization. We're candid about where the difficulty actually lives: the solver and the data.
The ~$60k/yr CRM. Building is cheap now; owning a daily-critical field app for 100+ reps is not. Leads with feature-scoping and total cost-to-own, not build time — and recommends replacing the highest-value module first.
High-value lead intelligence (geofenced dealers, missing-photo trends, priority list, weekly report to Ryan & Lisa) — separated honestly from the cold-email / scraping / list-purchase layer and its CAN-SPAM and deliverability landmines.
Smaller or underspecified items. A couple are quick once defined; others need a conversation before they're buildable.
A short discovery pass first, then ship one pilot while we line up the rest.
Start with the quick wins that replace clear manual labor or ride patterns we already run. Prove value fast, build trust, then move up the difficulty curve.
Almost everything good here depends on Predian (billing) and Rippling (hours, addresses), plus the photo/inventory source. That access is the critical path — worth confirming before we scope. For Predian a login may be enough if there's no MFA.
The First Photo checker is the strongest first pilot: concrete, measurable, AI-native. Land it, then redirect capacity onto the next tier as AutoTurn dev requests free up.
AI makes the build cheap. For systems like a custom Mobiwork, the real story is maintenance, support, and reliability for the people who depend on it daily. We say so plainly in each plan.