Hipteca

Evicron built for Hipteca, a Catalan mortgage advisory operating in the Maresme area, a custom AI-powered CRM that manages the full lifecycle of a mortgage operation: from the initial financing pre-assessment form all the way to signing. The system combines automatic digitisation of handwritten forms with Claude, a conversational assistant that queries the real database, and a calendar that genuinely syncs to the team’s phones — all behind a fully bilingual Catalan-Spanish interface.

The challenge

Hipteca worked the way many mortgage advisories do: the pre-assessment form was filled in by hand, on paper, during the client meeting. Each form holds around thirty critical data points — income, employment contract, CIRBE, bank, property details — that someone later had to transcribe. Sales follow-up lived scattered across calls, notes and memory, with no clear picture of where each operation stood. And since the firm operates in a bilingual environment, the tool had to feel equally native in Catalan and in Spanish.

The solution

We designed and built an internal, end-to-end management CRM with React, TypeScript and Firebase, running on 100% European cloud infrastructure (functions in europe-west1, data stored in the EU):

• Client management with automatic phone-number deduplication.
• Case files and pre-assessment forms supporting up to 2 holders and ~30 structured fields.
• A case state machine — follow-up → closing → completed, with sub-states — so the team always knows exactly where each operation stands.
• An immutable audit trail and real-time comments on every file.
• A calendar with 3 views (month, week and day), events linked to files and clients, an overdue drawer, and real phone sync via a subscribable iCal feed for Google and Apple Calendar, auto-refreshing roughly every 15 minutes.
• A metrics dashboard by state and 2 roles (admin and sales) with differentiated permissions and document-level security rules.
• A print view for every case file.

AI that saves hours of admin work

The differentiating piece of this project is how AI removes real administrative work rather than decorating the interface.

Digitising handwritten forms. Sales reps photograph the handwritten pre-assessment form and Claude (Anthropic), using vision and structured output, automatically extracts ~30 fields per form: holders, income, contract, CIRBE, bank, property details. What used to be manual transcription is now a few seconds of review. We also built a batch import to digitise the advisory’s entire paper archive.

A conversational assistant over real data. We integrated a chatbot that queries the CRM’s database in natural language, in Catalan or Spanish, through tool use: it lists and counts case files, clients and statistics. It doesn’t make data up — every answer comes from real queries against the system, with a visible monthly cost control for the team.

The result is a single place where the entire mortgage operation lives: fewer hours of admin, orderly sales follow-up, and an AI layer that works with the business’s data rather than around it.