Case study

BZ Solar Solutions

BZ Solar Solutions, a five-branch solar company in Mindanao, Philippines. Active client, running in production.

6 to 27sto a human on the first nine live chat requestsSmall sample, stated plainly: of the first nine requests through the live console, every one was accepted by an agent inside 27 seconds, and seven of the nine went on to get a reply. The two that did not are why the system flags an unanswered request instead of letting it disappear.

5branches on one pipelineQuote intake, lead records, and the pricing catalogue are shared, so an answer does not depend on which branch picked up.

0details required to see an estimateThe estimator runs before any contact capture. Phone number capture is separately consent gated rather than bundled into the same form.

The response times above are measured from the live chat console, not estimated. The sample is nine requests and is labelled as such rather than scaled into a percentage.

Where they started

Five branches, five different ways of handling an inquiry. A homeowner asking about a hybrid system got a different answer depending on who picked up, and the quotes that followed lived in whichever spreadsheet that branch happened to use. Nobody at the top could see the pipeline as one thing.

What it was costing them

Inconsistent answers cost trust, and a homeowner comparing three installers notices. Worse, an inquiry that arrived after hours had nowhere to go, so the lead either waited until morning or went to whoever replied first. In solar the second quote usually loses.

What I built

A public estimator anyone can use without handing over their details first, so the visitor gets value before being asked for anything. Behind it sits Mar, an AI assistant that answers common solar questions and qualifies the inquiry. Mar is deliberately not the last word: after a set number of answers it offers a handoff, and a real agent can join from a live console that already shows the full conversation, so the customer never has to repeat themselves. Everything lands in an invite-only staff workspace with role-based access, because the pipeline holds real customer records.

The pieces

Self-service estimator and published package catalogue, with no contact details required to see the initial recommendation.
Five-branch quote intake plus Mar, an AI assistant for common solar questions, lead qualification, and human handoff.
Private lead pipeline, contact records, quotation review, pricing catalogue, role-based access, AI training, and beta live-chat and inbox workflows.
The staff workspace is intentionally not linked here because it contains live customer and operational records behind invite-only authentication.

Next.jsReactTypeScriptSupabaseOpenRouterResendVercel

How the engagement works

Built and delivered in phases against a fixed project scope, with the staff workspace kept behind authentication because it holds live customer records.

Other case studies

DevSolutions Solar

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