Wednesday afternoon: another quote I didn’t want to send
On Wednesday afternoon, my design studio in Hangzhou received an inquiry from a client surnamed Chen: a TV cabinet for a 3.2-meter-wide living room, with room for a router and power strip. The budget was 6,000 yuan, but a brand-name store quoted nearly 20,000 yuan. My first reaction was to suggest the client scale back. Now I realize I got that wrong: customers aren’t buying a drawing. They’re buying confidence that the project is feasible and that they know what it will ultimately cost.
I’m not automating furniture production; I’m clarifying the idea first
So I used Claude Code to build a pre-communication tool for custom furniture. I enter wall dimensions, budget, storage items, and reference photos; it asks for missing information, then generates a structural sketch, material list, and questions to confirm. Designer Chen Lei used it with a client in a two-bedroom model apartment in Suzhou, turning “I think we could do this” into something both sides could discuss. It doesn’t sign drawings for carpenters or place orders automatically. It works well for early-stage alignment, not as a production-ready deliverable.
Replicating it today costs less than I expected
Money: about 200 yuan per month for a basic coding assistant. Time: two hours to build the input form, then three more to refine the questions and output format. Technical barrier: you can use the ready-made conversational version without coding, but turning it into a fixed form requires help setting it up once. First step: open Claude Code, create a project folder, have it read a “Client Requirements Form,” and ask only the five most critical questions. If setting that up feels like too much for now, manually paste the client’s information into the chat first and test whether the client will pay for a visual proposal. No pressure to try today.
How I’d choose based on where my business is now
If I were just starting and didn’t yet have a reliable way to find clients, I’d use my own bedroom or desk as the sample project and treat the proposal as a pre-quote communication aid, not rush to turn it into a product. If I already had one or two custom clients, I’d add version history, budget ranges, and material substitution options so every revision has a clear basis. If I were scaling up, I’d connect dimension checks, quote approvals, and carpenter confirmation forms while keeping human review in the loop; I wouldn’t send AI-generated dimensions straight to material cutting. Who’s using it matters less than one thing: whether clients will pay to “understand the proposal faster.”