# R-Szoft

> R-Szoft is a reviewed AI agency in Budapest, Hungary, listed in the AIAgencies.eu register and last assessed 2026-09-23.

Canonical page: https://www.aiagencies.eu/agency/r-szoft
Generated: 2026-09-28 · Curated by Gabor Kiss · No paid placements — inclusion, score, and rank are never sold.

## Facts
- Location: Budapest, Hungary
- Website: https://r-szoft.hu
- Hourly rate: €20-40/hr (directory-listed)
- Team size: 10-49
- Founded: 2018
- Services: Generative AI, AI Automation, AI Consulting
- Industries: SaaS & B2B, Manufacturing, E-commerce
- LinkedIn: https://www.linkedin.com/company/r-szoft

## About
R-Szoft is a Budapest software firm building generative-AI and agentic back-office automation for document processing, quotation, customer service and HR workflows, including on-premise open-model deployments.

## Documented clients
Documented engagements: Vodafone, Xella, and Indivizo. All 3 carry a source URL on file.

## Curator assessment
Reviewed 2026-09-23 by Gabor Kiss against published evidence.

R-Szoft's AI evidence is its own products plus six anonymized scenarios: IDP AI extracts Hungarian invoice fields with the open Qwen3-Vision model and reports an 86% straight-through processing rate, and RAITEST, a QA agent built for internal use, reports 90%+ of its generated tests succeeding against a live multi-portal SaaS application.

The case-study page credits unnamed clients with large results—a construction firm's quote preparation time down 75% and bid win rate up 35%, a manufacturer's first-time fix rate up 80%—but names no client, date or stack, so none of it can be checked.

Every case study that does name a client—Vodafone, Xella, Indivizo—describes conventional software work in Angular, Vaadin, Spring Boot or Go, with no AI component.

The weaker fit is buyers who need a named reference running an R-Szoft AI system, or who want the published percentages shown as measurements from a client deployment rather than scenario summaries.

### Key strengths
- Named enterprise delivery with technical detail: a recursive query engine and Angular front end for Vodafone, and a SAP-integrated B2B ordering platform on Vaadin and Spring Boot for Xella
- IDP AI publishes its stack and controls: Qwen3-Vision, Keycloak OIDC, audit logging, and per-field confidence scores that send only low-confidence fields to human review, deployable on-premise
- The RAITEST write-up explains its mechanics rather than asserting them: Jira-triggered test plans, Playwright exploration, a three-level self-healing locator, and local or cloud models

### Good to know
- None of the six AI case studies names a client—ask for one reference you can call whose AI system is still running, and how the 75% and 85% figures were measured
- The homepage says several large Hungarian companies use its ChatGPT alternative, AtheneAI, but names none—ask for a customer reference and which models it runs
- No team page names the engineers—ask who on your project has shipped LLM systems, since the named project history is Java, Angular and Go development

---
Source: AIAgencies.eu — the curated register of European AI agencies. Rate provenance: "agency-confirmed" = disclosed via submission or claimed profile; "directory-listed" = Clutch or two agreeing B2B directories (lone unverified sources are never written). Full method: https://www.aiagencies.eu/methodology · Rate index: https://www.aiagencies.eu/rates