Client: L. Speciaal Lasbedrijf — independent concept redesign, not affiliated with or endorsed by the business Location: Rotterdam-Kralingen, Netherlands Project Type: B2B industrial website rebuild + automated lead triage Timeline: One weekend, solo — Base44 #GiveItAGlow Challenge Technologies: Base44, Make.com, OpenAI (GPT-5 mini), Gmail Live website available at: https://lasbedrijf-speciaal-lasbedrijf.base44.app/

While most of the challenge was booking calendars for cafés and salons, I went looking for something higher stakes: a real B2B business where one person could say yes on the spot, no procurement committee. I found L. Speciaal Lasbedrijf — an industrial welding and metalworking company that's been running in Rotterdam-Kralingen since 1950, now in its second generation of family ownership, with over 20 real five-star reviews on Google.

Their contact page didn't have a broken form. It had no form at all — just a raw Contact Form 7 shortcode sitting exposed where a "Request a Quote" form should have been. For a B2B company, that's more than a visual bug: a prospective customer needs a clear way to explain the job, provide project details, and send supporting files. Every day it sat like that, a 76-year-old business with a real reputation was quietly losing quote requests from companies that needed an answer fast.

Even with a working form, every RFQ would still arrive as an undifferentiated request. A stainless-steel marine job and a routine repair need very different levels of attention. The better solution wasn't just collecting leads, it was actually capturing, classifying, and surfacing them automatically.

Rebuilt the contact experience as a multi-step Request a Quote flow: contact details -> material -> deadline -> quantity -> project description -> file uploads

Added real validation for email addresses and phone numbers

Reworked the capabilities and company pages around the company's documented history and business information.

The visual direction was deliberately industrial and structured, designed to feel like a serious B2B system rather than a template.

Added workflow logic so every incoming quote request is checked against material and description

Stainless steel, marine, and structural work is automatically surfaced as High Priority

Hit Base44's paywall around custom backend functions almost immediately, so I continued with the platform's native visual Workflows instead of stalling the build

Built a parallel pipeline: RFQ → Webhook → OpenAI → Gmail. Make.com owns classification and outbound communication; Base44 owns the application data and dashboard. That separation was intentional: 'two systems should never independently message the same client for the same event'.

Two real bugs turned up during testing, not staged ones: the AI step was reading placeholder text instead of the submitted form fields, and an email step was passing the raw AI response straight to the client instead of clean output. Both traced back to the same fix: mapping the real fields explicitly.

Built a secured /admin Kanban board : New → Analyzed → Quote Sent → Won → Lost.

Each RFQ opens into a detail view with contact info, material, deadline, quantity, project description, uploaded files, priority status, and a one-click email action. High-priority RFQs surface directly on the board, so nothing urgent gets buried in an inbox.

*Because this was an independent concept build, these are implementation results rather than measured business outcomes.*

Replaced the unusable contact experience with a structured RFQ flow

Real validation before submissions enter the workflow

File uploads for project documentation

Incoming RFQs classified automatically

High-priority jobs surfaced before manual review

A centralized RFQ pipeline instead of a basic inbox

Full RFQ detail views with uploaded-file access

Client confirmations and internal alerts triggered automatically as part of the workflow

*I called and emailed the real owners the Friday this was built. They were closed for the weekend, so as of publishing, I'm still waiting to hear back. I'd rather say that plainly than fabricate a response. If that changes, I'll update this with what actually happened.*

Base44: Front end, RFQ flow, data, admin dashboard, native workflows Make.com: Automation orchestration, AI classification, outbound email OpenAI (GPT-5 mini): RFQ priority classification Gmail: Client confirmations and internal alerts

Architecture: Customer → Base44 RFQ → Make.com → OpenAI → Priority Classification → Gmail, with the Base44 Admin Dashboard reading from the same data layer.

Base44 owns the application and data. Make.com owns automation and communication. OpenAI handles classification. Each system has one clearly defined job, that split is the actual design decision this project is about.

This started as a website challenge and turned into an exercise in designing the workflow behind a B2B site.

The real lesson: a quote form is only the first step. The useful system is the chain that follows: capture, classify, prioritize, notify, manage.

For potential clients: if your quote or contact form has been broken for even a week, you're losing exactly this kind of lead, the ones who won't email twice. I build the whole system: the front end, the data model, and the automation that decides what happens the moment a submission lands.

For fellow builders: when a platform paywalls the automation layer you need, build it as a parallel system rather than forcing everything through one tool, but pick a single source of truth immediately, before two systems both try to email the same person.