Lead Capture already lives in three places. Why not a fourth?
Lead Capture already runs on IHSS devices, in the PROvider app, and is currently extending to associates in-store. Each of those puts an actual person, or a device an associate is holding, between the customer and the request. This prototype asks what happens if lowes.com itself became an entry point: a customer researching a water heater at 9pm, with no associate or provider in the loop yet, still ends up with a qualified request in the system. It's also a more automated, effortless way to generate leads in the first place — the request still lands in the same Lead Management System this whole ecosystem already runs on, just generated without anyone needing to be there to capture it.
MyLow: answers first, then a request that writes itself
MyLow is a conversational AI assistant embedded on lowes.com, scoped to a specific product category — in this prototype, water heaters. A customer can pick from suggested questions or type their own, and MyLow answers directly, in plain language, before ever asking for anything. Once real purchase intent shows up — the customer asks about installation, pricing, or timing — MyLow offers to request a personalized quote. The goal was never just to answer a question: it's to connect the customer with a Lowe's professional who, by the time they reach out, already has the full context of what was discussed, not just a name and a phone number. If the customer says yes, MyLow assembles a structured lead from the conversation itself: project type, what was discussed, and contact details, ready to review and submit. No form, because the conversation already was the form.
Where the AI actually helps
Typing a free-form question doesn't just return a static answer. Here, "How much does a replacement water heater cost?" gets a real answer, then MyLow asks what's actually going on — a broken unit or a planned upgrade — and follows with specific, branching questions about fuel type and current unit. Each answer narrows what MyLow understands about the job, without ever feeling like a form.
The request MyLow builds isn't generic. It reflects the specific path this conversation took: "Existing electric, tankless water heater — unit no longer working," pre-filled contact details, and a clear next step. The customer reviews it once and submits — nothing to re-type, nothing generic about what gets sent to the provider.
Further than a mockup, short of a real product
Unlike a static flow diagram, this one actually works: the chat responses, the shift from answering to offering a quote, and the AI-assembled request summary all run live in the prototype, not just as connected screens. What it doesn't have is a real backend, real integration with Lowe's lead systems, or any customer in front of it yet. It's a working demonstration of the idea, not a shipped feature.
Why show unshipped work at all
Most of what's on this site is proven — shipped, piloting, or backed by real numbers. This isn't, and I want that to be obvious rather than blurred. I'm including it because craft and judgment show up before something ships too: in noticing that Lead Capture already has three live entry points and asking whether a fourth, conversational one earns its place, in deciding where an AI should answer versus where it should just get out of the way and hand off to a human. This is what that thinking looks like before it becomes a real project.