Case study

Magistral CRM

I designed, built and run Magistral Constructora's CRM: leads from their Meta ads come in on their own, get assigned to the sales team automatically, and each one keeps its full history until the sale.

Client: Magistral Constructora
My role: Everything: product, design, development and operations
Period: February – October 2026, in production
Live: magistralcrm.pro

Leads table with stages, project, salesperson and attempts

+3 000

Meta leads processed

forms resolved automatically

230

commits in 8 months

The problem

Ad leads arrived in one place and follow-up happened somewhere else: spreadsheets, WhatsApp and memory.

Assigning them was manual, nobody knew how many times each client had been called, and some got lost with no explanation.

And the usual question had no answer: which campaign sells and which one only spends?

What I built

Automatic ingestion from Meta

Every form submission arrives by webhook and becomes a lead. If a marketer words the questions their own way, the system works out which one is the phone and which the name in layers: Meta's question type, known names and, last, AI validated against the answer.

Assignment and follow-up

Automatic assignment by rotation, form, ad or percentages. Each salesperson gets the lead on Telegram with everything needed to call, and the pipeline, reminders and contact policy handle the rest.

Reconciliation and analytics

What Meta sent versus what entered the CRM, submission by submission, plus spend and cost per lead for each campaign from the Graph API. No lead is lost without an explanation.

Roles, permissions and audit

Roles with granular capabilities, impersonation for support and a log of every change with its before and after.

Built with

Next.js 16React 19ConvexTypeScriptTailwindshadcn/uiOpenAIMeta Graph APITelegram Bot APIClerk

What it looks like

Meta forms resolved with AI — Forms resolved in layers, with the AI's explanation.

A lead's detail panel — The detail panel: stage, lots of interest, answers and to-dos.

Campaign summary with spend and cost per lead — Which campaign sells and which one only spends.

Notifications on a salesperson's phone — How each lead reaches the salesperson.

What came next

The system became a product: edercrm takes what worked at Magistral, simplifies it for small teams and adds a lot inventory. The same happened with the quoting tool I built for their showroom, which is now eder-cotizador. The screenshots on this page come from edercrm, with made-up data.

edercrm, the product version

Does your company need something like this?

Message me on WhatsApp