MoveCubic: from enquiry to a quote movers can trust

The challenge

Moving companies lose hours on every quote. Details arrive by phone, form and email; someone pieces the job together; inventories and tariffs are copied between tools before a price can be checked. Quick descriptions miss items and access details, which leads to under-quoted jobs or a site visit for every enquiry.

The approach

We designed MoveCubic around the estimator, not around the AI. Customers send room-by-room photos; the system proposes an inventory with source photos and uncertainty flags; confirmed items become a volume range; and the company's own versioned rate card produces a transparent draft price. Nothing is sent to a customer until an estimator approves it.

What gets built

• A TypeScript modular monolith: Next.js for the marketing site and tenant app, NestJS for the API and a separate background worker sharing one codebase.
• Multi-tenant data model in MySQL with Prisma, where every query is scoped by a server-derived tenant context and covered by cross-tenant regression tests.
• AI inventory extraction behind provider adapters, returning validated structured output with uncertainty flags; model, prompt version, latency and cost are recorded for every run.
• Pricing is a server-side formula on versioned rate cards using exact money arithmetic, so a later tariff change can never silently alter an accepted quote.
• Redis and BullMQ for background jobs, a transactional outbox for side effects, and idempotent jobs and webhooks so a retry never sends a quote twice.
• A bilingual (Dutch and English) marketing site with a server-validated pilot enquiry flow, built to WCAG basics and reduced-motion support.

Outcome

MoveCubic is in late development and launching soon with a founding pilot for Dutch moving companies. We will publish measured results, such as quote turnaround time and estimator correction rates, once pilot customers have used it on real jobs.

Technologies