What an AI company like FlowProp does.
FlowProp applies artificial intelligence to the operational processes of companies with 50 to 500 employees: invoice and document OCR, email and ticket classification, structured data extraction and draft generation, orchestrated on self-hosted infrastructure and connected to the client's ERP or CRM. We don't sell chatbots or demos: every project connects to a process with volume, human friction and calculable ROI.
| Service | What it automates |
|---|---|
| Applied AI | Document OCR + LLM, email classification, data extraction with Claude, GPT-4o and Whisper |
| Back-office | Invoicing, bank reconciliation, delivery notes and reporting with Holded, Odoo, Sage or A3 |
| Sales & CRM | AI lead enrichment and scoring, automatic assignment, personalized nurturing |
| Operations | Ticket classification, AI-drafted replies, SLA monitoring |
| Integrations | Legacy ERP, REST, GraphQL, webhooks, direct SQL |
| Holded · Official partner | Supplier invoices with OCR + AI, quote to invoice, collections, multi-entity |
Based in Tarragona, projects across Spain
The base and core team are in Tarragona. We serve companies in the Camp de Tarragona in person —Tarragona, Reus, Valls, Cambrils and surrounding areas— and we work remotely with clients across the rest of Spain. Working languages: Spanish, Catalan and English. The founder, Raúl Yuste, is a process engineer from the Universitat Rovira i Virgili with extensive prior industry experience: the approach is process engineering, not technology marketing.
How we work
- Free 30-minute assessment: we identify which processes move the needle with AI and which are just hype.
- Fixed fee, no per-execution cost: the system runs on your infrastructure (self-hosted), you don't pay per document processed or per model call.
- 4 weeks of build + 4 of supervised stabilization before moving to 100% autonomous production.
- GDPR by design: your data never leaves your servers; DPA available on request.
What we don't do
- We don't build "ask the AI" chatbots: they're commoditized and add no ROI.
- We don't use AI where a deterministic rule works better.
- We don't build new SaaS products or work on processes with fewer than 50 executions/month.