Case study

DevSolutions Solar

DevSolutions Solar, a solar installer in Cavite, Philippines. Active client, running in production.

View the live system Read the build journal

2,066leads captured in the live systemCounted in the production CRM, not estimated.

754branded quotations generatedGenerated, not all delivered: 18 failed on an attachment timeout that has since been fixed, and 17 are queued for manual retry. The system now records both numbers so a gap like that cannot hide again.

212knowledge chunks behind every Sol answerIngested from 9 vetted company sources, so answers come from approved material rather than a model's guess.

18golden test cases run before every prompt changeDrawn from real customer transcripts. A prompt edit that breaks a known-good answer is caught before it reaches a customer.

Every figure on this page is read from the live production system or from the project's own build records. None of them are estimates.

Where they started

Inquiries arrived through Facebook comments, Messenger threads, and phone calls. Every quote was assembled by hand in a spreadsheet, then retyped into a document. Nobody could say how many leads were in play this week, and the same five product questions consumed the sales team every single day.

What it was costing them

A homeowner asking about a 6kW hybrid at 9pm got an answer the next afternoon, if anyone remembered. Quotes took days, so buyers who were ready shopped elsewhere while they waited. And because the pricing lived in one person's head and one spreadsheet, every quote carried the risk of a number nobody could later defend.

What I built

One connected system, marketing site through internal tooling. A homeowner sizes their system on a public calculator and gets an instant estimate. Their exact address is geocoded against the Silang warehouse, so logistics is priced by real road distance rather than a guessed zone. The lead lands as a card in a pipeline the whole team can see. Sol, an AI assistant grounded in the company's real product knowledge, answers pricing and sizing questions around the clock and hands off when it should. Approved quotes generate a branded PDF and email themselves out.

The pieces

Custom quotation calculator. Homeowners size their system, get an instant estimate, and download a branded PDF quote.
Lead capture forms wired directly into an internal pipeline (no CSV exports, no Zapier middleware).
Custom pipeline dashboard for the sales team. Every form submission lands as a card with stage, source, and follow-up history.
RAG chatbot for product knowledge, built on n8n + Supabase. Answers pricing, sizing, and policy questions in real time.

Next.jsSupabaseVerceln8nGitHub

How the engagement works

Built in phases and maintained on an ongoing monthly retainer, with each significant new capability quoted separately.

Other case studies

BZ Solar Solutions

Your bottleneck is probably not solar.

The pattern travels: capture the lead, answer them before they cool, and put the follow-up somewhere it cannot be forgotten. Tell me where yours leaks and I will tell you straight whether it is worth automating.

Start a conversation