> Markdown version of [/videos/1311-graphs-and-rags-everywhere-but-what-are-they-andreas-kollegger-neo4j?t=1](https://www.wearedevelopers.com/videos/1311-graphs-and-rags-everywhere-but-what-are-they-andreas-kollegger-neo4j?t=1). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md).

---

# Graphs and RAGs Everywhere... But What Are They? - Andreas Kollegger - Neo4j

Andreas Kollegger argues that AI agents shouldn't be autonomous black boxes. Discover how combining GraphRAG with local LLMs turns agents into safe, composable software design patterns.

- **Speakers:** 
- **Event:** 
- **Published:** March 19, 2025
- **Duration:** 46:51
- **URL:** https://www.wearedevelopers.com/videos/1311-graphs-and-rags-everywhere-but-what-are-they-andreas-kollegger-neo4j

## Summary

The rapid adoption of generative AI has exposed the limitations of relying solely on baseline language models, driving the need for domain-specific context via retrieval-augmented generation (RAG). While traditional vector similarity search provides a fuzzy lookup for unstructured data, it frequently fails to capture nuanced context or multi-hop relationships. To solve this, GraphRAG integrates knowledge graphs into the retrieval process. By structurally mapping both structured business records and distilled facts from unstructured files into native nodes and relationships, graph databases eliminate expensive cross-table joins while supplying LLMs with a highly focused signal-to-noise ratio.

Building this architecture does not require strict upfront schemas. Engineering teams can ingest raw datasets iteratively, creating a minimum viable graph that evolves as data requirements grow. This incremental linking immediately unlocks emergent analytical insights even before complex ontologies are established. Furthermore, as data privacy and proprietary governance become critical business priorities, the industry is pivoting toward smaller, highly capable open-source models. Running small or local language models allows organizations to process sensitive context continuously without exposing proprietary information to centralized cloud providers.

A significant paradigm shift occurs when developers reframe how they interact with AI agents. Rather than viewing agents as autonomous black boxes taking control of local resources, developers should treat them as the composable unit of LLMs. Approaching agents as a standard software design pattern allows teams to apply traditional access privileges, encapsulate side effects, and monitor reasoning traces just like any routine function call. By starting small and incrementally applying graph pattern matching directly to application logic, developers can abstract away boilerplate code and safely integrate agentic workflows into enterprise backend systems.

**Keywords:** graph database architecture, retrieval-augmented generation, knowledge graph data modeling, vector similarity search limitations, unstructured data ingestion, minimum viable graph, composable AI agents, open-source LLM deployment, local language models, generative AI data privacy, LLM signal-to-noise ratio, cross-table database queries, AI software design patterns, agentic memory structures

## Chapters

1. **Introduction to Neo4j and remote developer relations work** (00:01) — An overview of Neo4j's growth and the global travel patterns of their developer relations team.
1. **Understanding graph databases and node relationship modeling** (03:32) — How graph databases use nodes and relationships to connect and organize data efficiently without massive join tables.
1. **The fundamentals of retrieval augmented generation operations** (06:05) — How preparing language model prompts with external retrieved context guarantees more accurate operational answers.
1. **Enhancing language models with graph retrieval augmented generation** (08:58) — Using knowledge graphs to distill information and provide highly focused context for answering language model queries.
1. **Merging structured and unstructured data for business insights** (11:24) — Techniques for connecting fragmented text documents with existing business data records to build comprehensive query systems.
1. **Resources and introductory tools for learning graph technologies** (12:33) — Available online courses and bundled functional packages that assist developers with implementing graph knowledge representations.
1. **Handling messy enterprise data with evolvable graph schemas** (14:03) — Mapping fragmented spreadsheets and legacy documentation into incrementally refined graphs simplifies massive enterprise data migrations.
1. **Returning to basic text formats for machine learning** (18:21) — Why straightforward text formats like Markdown and CSV are preferred for feeding clean unstructured data directly to automated AI scrapers.
1. **Addressing data privacy concerns with local language models** (19:14) — Running self-hosted language models ensures corporate data governance and prevents critical proprietary information from leaking.
1. **The impact of open source models on industry dynamics** (20:54) — How advanced open source releases disrupt traditional funding incentives and democratize advanced artificial intelligence capabilities.
1. **Overcoming language model limitations via streaming data ingestion** (23:47) — Processing real-time information updates with specialized small models maintains current contextual relevance without high resource consumption.
1. **Embedding local language models into user operating systems** (26:21) — How small models operating as personal desktop agents enable advanced contextual search and independent file interactions.
1. **Treating artificial intelligence agents as composable software units** (28:18) — Reframing machine learning agents from autonomous actors into controllable software architectures managed by traditional design patterns.
1. **Comparing graph augmented retrieval with vector similarity search** (33:36) — Why basic vector matching lacks the specific contextual relationships that knowledge graph cross-linking effectively secures.
1. **Overcoming developer intimidation when integrating foundational graph structures** (37:20) — Recognizing that functional graph querying relies heavily on accessible pattern matching approaches rather than entirely native data components.
1. **The future potential for native graph programming languages** (41:11) — Exploring how contemporary coding frameworks abstract internal graph operations and the viable opportunity for newly tailored native syntax.
1. **Elevating developer workflows by automating repetitive foundational tasks** (43:24) — How automated tools securely shoulder repetitive dependency tracking constraints so programmers can target impactful software challenges.
1. **Implementing sensible adoption strategies for generative programming systems** (45:05) — Applying measured incremental implementation strategies lets teams confidently adopt emerging capabilities without enduring overwhelming complexity.

## Related Moments

- [Introduction to generative AI and knowledge graphs](https://www.wearedevelopers.com/videos/1154-large-language-models-knowledge-graphs) (from "Large Language Models ❤️ Knowledge Graphs")
- [Unlocking generative AI capabilities using knowledge graphs](https://www.wearedevelopers.com/videos/100318-context-graphs-for-explainable-decision-aware-ai-agents) (from "Context Graphs for Explainable, Decision-Aware AI Agents")
- [Protecting enterprise data with local models and RAG](https://www.wearedevelopers.com/videos/1098-decoding-trends-strategies-for-success-in-the-evolving-digital-domain) (from "Decoding Trends: Strategies for Success in the Evolving Digital Domain")
- [Transforming unstructured email data into customized user knowledge graphs](https://www.wearedevelopers.com/videos/100128-the-golden-age-of-email-owning-the-inbox-in-the-age-of-ai) (from "The Golden Age of Email: Owning the Inbox in the Age of AI")
- [Automating structural graph generation using large language models](https://www.wearedevelopers.com/videos/1543-from-tables-to-graphs-in-minutes-supercharging-kusto-graph-analytics-with-ai-powered-development) (from "From Tables to Graphs in Minutes: Supercharging Kusto Graph Analytics with AI-Powered Development")
- [Building local RAG architectures using the Anything LLM tool](https://www.wearedevelopers.com/videos/1597-self-hosted-llms-from-zero-to-inference) (from "Self-Hosted LLMs: From Zero to Inference")

## Related Articles

- [Graph and AI Trends 2026: Why Is AI Running but Not Yet Delivering?](https://www.wearedevelopers.com/magazine/680-graph-and-ai-trends-2026-why-is-ai-running-but-not-yet-delivering)
- [Everything a Developer Needs to Know About MCP with Neo4j](https://www.wearedevelopers.com/magazine/604-everything-a-developer-needs-to-know-about-mcp-with-neo4j)
- [MLOps And AI Driven Development](https://www.wearedevelopers.com/magazine/82-mlops-and-ai-driven-development)
- [What is Agentic Programming and Why Should Developers Care?](https://www.wearedevelopers.com/magazine/625-what-is-agentic-programming-and-why-should-developers-care)

## Related Jobs

- [AI Software Engineer (Germany)](https://www.wearedevelopers.com/jobs/48317-ai-software-engineer-germany) at **Sunhat**
- [Software-Entwickler – RAG & Knowledgraph (m/w/d)](https://www.wearedevelopers.com/jobs/48330-software-entwickler-rag-knowledgraph-m-w-d) at **Riverty**
- [AI & Machine Learning Engineer (all genders)](https://www.wearedevelopers.com/jobs/48217-ai-machine-learning-engineer-all-genders) at **msg**
- [Data Scientist](https://www.wearedevelopers.com/jobs/ext/1351648-data-scientist) at **Almedia**
- [Principal Software Engineer, Enterprise AI Platform](https://www.wearedevelopers.com/jobs/ext/1467292-principal-software-engineer-enterprise-ai-platform) at **GitHub**
- [Senior AI Agent Software Engineer (Go, Python) (m/f/x)](https://www.wearedevelopers.com/jobs/48277-senior-ai-agent-software-engineer-go-python-m-f-x) at **Dynatrace**