Customer reference
Company Brain for Tracify: GBrain Running Locally, EU-Sovereign, Built with Claude Code
Tracify GmbH · MarTech / marketing attribution · How Pexon Consulting built a company brain on the open-source GBrain layer, connected every data source, and hosted it EU-compliantly on AWS Bedrock so Tracify's CEO and management query synthesized, cited answers instead of re-reading five systems.
Running a company brain in production means routing every model call through one place, with budgets, keys and audit trails attached. That layer is what our LiteLLM gateway consulting covers, from routing and cost tracking to GDPR-compliant operations.
Updated: July 2026
Key facts at a glance
Who
Tracify GmbH, a German marketing-attribution SaaS used by performance marketers steering over €1 billion in ad budgets, with infrastructure in Munich and Frankfurt.
Problem
Institutional knowledge fragmented across docs, chats, tickets, decks and dashboards, while generic search returned pages instead of answers and GDPR blocked uncontrolled US model endpoints.
Solution
Pexon Consulting built a company brain on GBrain, connected all relevant data sources, and delivered EU-compliant Claude access via AWS Bedrock with pgvector hybrid multi-query retrieval and a self-wiring knowledge graph.
Result
Tracify's CEO and management now query one synthesized, cited source of truth in production, preserving decision context as the company grows without DSGVO risk from arbitrary external endpoints.
Results
€1B+
Ad budgets steered · Tracify attribution platform at scale
All
Data sources · Relevant company systems connected to one brain
EU
Data residency · Claude via AWS Bedrock, GDPR-aligned
Live
Leadership adoption · CEO and management in daily production use
Client overview
Industry
Marketing attribution / MarTech SaaS
Company
Tracify GmbH
Region
Germany (Munich and Frankfurt)
Services
Company brain architecture, EU-compliant AI platform, data-source integration
Tech stack
GBrain, AWS Bedrock, Claude, Claude Code, Claude Cowork, PostgreSQL, pgvector
Project duration
End-to-end build engagement
Compliance
GDPR (DSGVO), EU-compliant AWS Bedrock boundary
Status
Company brain in production use by CEO and management
Starting point
Fragmented institutional memory
Docs, chats, tickets, decks and dashboards each held part of the story. The answer to what was decided and why lived in five places, and reconstructing it cost leadership the most time.
Search returns pages, not answers
Generic search tools surfaced a list of hits leadership still had to open and read. What executives need is a synthesized, cited answer with an honest note on what remains unknown.
Lost context as the team grows
Past decisions, deals and rationale faded as headcount increased. A fast-growing SaaS selling data accuracy faced the same internal knowledge problem it helps customers solve.
GDPR constraint on AI adoption
A German SaaS handling customer-adjacent data cannot route everything through arbitrary US model endpoints. AI usage had to stay DSGVO-compliant with controlled, in-region data flow.
Solution & approach
1
GBrain synthesis and knowledge graph
Pexon built on GBrain instead of a bare vector store: a synthesis layer returns cited answers with explicit gap notes, and a self-wiring knowledge graph creates typed edges between people, companies and topics on every write.
2
All data sources connected
Every relevant Tracify system feeds one brain so institutional memory is consolidated. Leadership queries across the full corpus instead of hunting through disconnected tools.
3
EU-compliant Claude on AWS Bedrock
Claude is consumed through AWS Bedrock inside an EU-compliant boundary. Model traffic and data stay in-region for DSGVO conformity instead of flowing to unmanaged public endpoints.
4
pgvector hybrid multi-query retrieval
Retrieval fuses vector similarity via pgvector with keyword matching and multi-query expansion, surfacing results on their strongest evidence across the corpus. Delivery used Claude Code and Claude Cowork end to end.
Before / after
| Dimension | Before | After | Impact |
|---|---|---|---|
| Knowledge access | Five disconnected tools | One cited source of truth | Less executive re-reading |
| Search quality | Page lists to open manually | Synthesized answers with citations | Answers, not hits |
| AI compliance | GDPR risk from open endpoints | Claude via EU AWS Bedrock | Controlled in-region flows |
| Decision context | Rationale lost as team grows | Preserved institutional memory | Leadership-grade adoption |
“
We now query one synthesized, cited source of truth instead of re-reading five systems. Pexon connected our data sources on GBrain and kept AI usage inside an EU-compliant AWS Bedrock boundary.
TM
Management, Tracify GmbH
Leadership · Marketing attribution SaaS, Germany
Technologies & services
GBrain
AWS Bedrock
Claude
pgvector
Claude Code
Knowledge graph
Frequently asked questions
What is a company brain, and how is it different from search?
Search returns pages you still have to read. A company brain reads them for you and returns a synthesized, cited answer, and flags what it does not yet know. Pexon built Tracify's on the open-source GBrain layer with synthesis and a self-wiring knowledge graph on top of retrieval.
How is the AI kept GDPR-compliant?
Claude is accessed through AWS Bedrock in an EU-compliant boundary, so model traffic and data stay in-region instead of flowing to arbitrary external endpoints. That gives Tracify controlled AI adoption under DSGVO without bypassing governance.
Why GBrain instead of a plain vector database?
A bare vector store returns semantically close chunks. GBrain adds a synthesis layer with cited answers and gap analysis, plus a self-wiring knowledge graph that connects facts vector search cannot reach. On its published corpus, GBrain's retrieval layer benchmarks at P@5 49.1% and R@5 97.9%, the tool's published numbers, not a Tracify-specific measurement.
What does pgvector hybrid multi-query search mean?
Retrieval combines vector similarity via the pgvector Postgres extension with keyword matching and expands each query into several variants, then fuses the results. That yields higher recall than vector-only retrieval across Tracify's full connected corpus.
Who actually uses the brain?
Tracify's CEO and management use it in production to support decisions, not just the engineering team. Pexon designed for leadership-grade daily adoption: cited answers leadership can trust, with explicit notes when the brain lacks evidence.
Related case studies
AI education
Custom ERP on AWS Bedrock
Claude · EU Bedrock AZAV-compliant platform with governed AI
Read case study
ManufacturingRAG pricing and voicebot
RAG · Voice AI on proprietary company knowledge
Read case study
ManufacturingClaude enablement EU-compliant
Claude Code · EU Data Zone Hands-on AI training for management and engineering
Read case study
Institutional memory as a cited company brain
Ready to consolidate scattered knowledge into one EU-compliant brain with synthesized, cited answers for leadership? Pexon Consulting covers architecture, platform and integration in one engagement.
Sie suchen einen Partner für Ihr Projekt?
Wir analysieren Ihren individuellen Bedarf, geben erste Empfehlungen zu Umsetzungsstrategien und bieten Ihnen transparente Einblicke in unsere Methoden, Technologien und Referenzen.
400+
Projekte seit 2020
100+
Kunden im DACH-Raum
ISO
27001 zertifiziert
3
Hyperscaler Partner