AI development from Frankfurt

We build your
AI solution

Custom GPTs with RAG, API integrations into existing systems, document AI pipelines, AI web apps. Fixed price, GDPR-compliant, hosted in Frankfurt. Built for German SMEs — by intercolo GmbH.

Recommended starting point

Already know what you want? Let's get started.

If you already know exactly which AI functionality you need — pick a package below and we'll get going. If you're still unsure what makes the most sense for you, start with our AI consulting (AI potential check from €2,900). ~40 % of our consulting clients then move on to development — with a clear, prioritised use case.

View AI consulting first
Fixed-price packages

Four ways to bring AI into your software

Clear packages, transparent prices, documented scope. More complex projects: individual fixed-price quote after discovery.

Popular

Custom GPT with RAG

A knowledge assistant on your own documents

from €4,900 net · 3–4 weeks

We build your Custom GPT with access to your documents (PDFs, DOCX, Confluence export). Web UI or embedded in EuGPT. Includes an admin interface for document updates.

  • RAG with vector database (Qdrant / pgvector)
  • Web UI or EuGPT embed
  • Admin for document updates
  • Source citations in every answer

Not included: SSO, multi-tenancy, live sync to source systems, mobile app.

Request Custom GPT

API integration

AI in your CRM, ERP, ticketing

from €6,500 net · 4–6 weeks

AI functionality built into your existing software: auto-classification, response suggestions, summaries. Includes API integration, logging and fallback logic.

  • HubSpot / Salesforce / Zoho / SAP / JTL
  • Shopware / Zammad / OTRS / JIRA
  • REST/webhooks, logging, fallback
  • EuGPT, Mistral, OpenAI — depending on requirements

Not included: Licence migrations, customisation of the source system, on-premise deployment.

Request API integration

Document AI

Receipt and contract processing

from €8,900 net · 5–7 weeks

Pipeline for invoices, contracts, delivery notes: OCR + structured extraction + validation + handover to the target system. Includes a review UI for edge cases.

  • OCR (Mistral / Azure Document Intelligence)
  • Structured extraction (JSON)
  • Review UI for validation
  • Hosted in Frankfurt, GDPR-compliant

Not included: DATEV certification, workflow engine, mobile capture app.

Request document AI

AI web app (MVP)

A complete application from auth to deployment

from €14,500 net · 6–10 weeks

A complete web application: authentication, core AI functionality, data model, admin area, deployment on Frankfurt servers. For internal tools or customer-facing MVPs.

  • Laravel + Tailwind or Next.js
  • Auth, data model, admin, deployment
  • Postgres, hosted in Frankfurt
  • Fixed change-request process

Not included: App Store apps, complex permission systems, multilingual, SOC 2.

Book an intro call

B2B prices are net, plus 19 % VAT; consumers receive a final price including statutory VAT before the contract is concluded. “From” means: fixed price after the requirements workshop. More complex projects: individual quote.

After go-live

Maintenance & support

Your software is live — now it's about stable operation and continuous improvement.

Package Price/month Scope Response time
Basic €290 Hosting monitoring, security updates, 1h of changes 2 business days
Standard €690 + 4h development/month, quarterly review 1 business day
Active · popular €1,490 + 10h development, monthly call, backlog maintenance 4h business day
Hourly €145/h Ad hoc, 4h minimum commitment, no SLA subject to availability

12-month minimum term, cancellable monthly thereafter. No 24/7 support, no weekend pager. With a small team, we're upfront about that — instead of promises we can't keep.

Example cases

What we typically build

Two anonymised build cases from the recent past — concrete, with figures and realistic effort.

Tax advisory firm, Frankfurt 25 employees

Custom GPT with RAG for BMF circulars

Problem

Tax advisors lost an average of 45 minutes per client case researching BMF circulars, application decrees and specialist literature. Their own PDF collection had grown over years, unstructured and unsearchable.

Solution

Custom GPT with a RAG pipeline (Qdrant) on 1.800 BMF circulars and internal notes. Answers include source citations with file reference numbers. Self-hosted in Frankfurt due to client confidentiality (§ 57 StBerG).

Result

BMF research cut from 45 minutes to 3 minutes — around 12 hours per advisor recovered every month. Source citations accepted by the professional association.

Tools EuGPT + Custom GPT with RAG
Duration 3 weeks of implementation

Anonymised example — real references available on request.

B2B software provider 120 employees

API integration into HubSpot CRM — lead triage & first responses

Problem

The sales team manually processed around 80 inbound leads a day — classification (ICP fit), prioritisation and the first response took 8–12 minutes per lead. Hot leads got lost in the volume.

Solution

API integration into HubSpot via webhook: the AI classifies each lead against ICP criteria, scores priority and suggests a personalised first response. The sales rep validates it with one click. Implemented vendor-neutral — the customer was already using ChatGPT Enterprise.

Result

12 hours/week saved across the sales team, response time on hot leads cut from 4h to 15 min. ROI reached in just under 4 months.

Tools ChatGPT Enterprise via API + HubSpot webhooks
Duration 5 weeks of implementation
Package API integration

Anonymised example — real references available on request.

Both cases are anonymised; concrete reference customers available on request. Effort and time savings vary depending on data quality and existing systems.

Process

How your software gets built

A clear process, clear responsibilities, no waterfall — but no scope creep either.

01

Requirements workshop

We understand your use cases and define scope and acceptable trade-offs. Fixed-price quote within 48h afterwards.

02

Prototype

A prototype is ready for validation within 1–2 weeks. Iterative feedback instead of a big bang.

03

Implementation

Agile development of the production-ready solution. API integration, tests, security review.

04

Go-Live & Hypercare

Deployment on German infrastructure. 30 days of Hypercare support included. Optional maintenance retainer afterwards.

Technology

What we typically use

A proven stack — no hype, no experimental projects at your expense.

# intercolo Development Stack
Backend=Laravel 12 (PHP 8.2+) / Python FastAPI
Frontend=Alpine.js + Tailwind / Next.js for SPA needs
RAG   =Qdrant, pgvector (Postgres-native)
LLMs  =EuGPT, Mistral, OpenAI — depending on compliance requirements
OCR   =Mistral OCR, Azure Document Intelligence
DB    =PostgreSQL, Redis
Hosting=GPU cluster and data centre Frankfurt (intercolo)
Fixed price · Frankfurt

Ready to bring AI into your software?

Choose a package or book a 30-minute intro call — we'll clarify scope and deliver a fixed-price quote within 48h.