Large language models (LLMs), with their capability to understand written text and generate human-like responses, are rapidly becoming indispensable across many sectors. Even intelligence analysts, with their security-conscious approach, are exploring the ways in which LLMs can enhance the analytical process.
Integrating link analysis with large language model (LLM) capabilities is a natural progression.
Law enforcement agencies are eager to take advantage of these developments. In fact, 80 percent of the papers submitted to leading police conferences now focus on large language models and AI.
Speakers from Europol and the National Policing Institute are actively sharing their approaches to using LLMs. Interpol in partnership with the United Nations Interregional Crime and Justice Research Institute (UNICRI) has released the Toolkit for Responsible AI Innovation in Law Enforcement, providing a comprehensive framework to help agencies adopt AI technologies responsibly and ethically.
Integrating link analysis with large language model (LLM) capabilities is a natural progression. The synergy between knowledge graphs and LLMs is particularly compelling: knowledge graphs provide structured, domain-specific context that grounds LLM outputs and significantly improves accuracy and explainability.
Intelligence analysis platforms built with graph technology, such as GraphAware’s Hume, exemplify this integration. By combining the contextual precision of knowledge graphs with the generative power of LLMs, Hume delivers a robust framework for intelligence analyst work.
GraphAware’s R&D experts have now drawn on their deep expertise in graph data science, machine learning, and AI to deliver a groundbreaking capability within the Hume platform: Maestro, the first chatbot built specifically for intelligence analysts.
GraphAware Hume Maestro allows analysts to interact with complex knowledge graphs using natural language instead of Cypher queries, dramatically lowering the technical barrier to powerful graph exploration. Whether generating code snippets, summarising key facts, or offering contextual guidance while navigating Hume, Maestro unlocks new advancements in intelligence analysis.
How LLMs transform front-line intelligence access
LLMs help analysts get a quick orientation in the vast amount of data available about a given case. Let’s take the example of an ultimate beneficial owner analysis conducted for the purpose of criminal asset confiscation. With an LLM-powered chatbot, analysts facing a complex knowledge graph containing information on tens of thousands of companies, their transactions and relationships, can easily ask the graph: “Which companies are ultimately owned by person XY?”, “Are there any potentially fraudulent transactions between these companies?” and get a relevant answer. The LLM, guided by the knowledge graph’s schema, converts these natural queries into precise Cypher commands that search the curated intelligence database.
All of these take just seconds to minutes in the graph of hundreds of thousands of companies and transactions.This approach has the power to save significant time exploring complex UBO fraud cases.
What makes this approach particularly powerful for law enforcement is its reliability. Unlike standard chatbots, LLMs in this use case are used only to translate questions into queries. Therefore, the information returned is always based on verified data within the knowledge graph. Officers can easily check that their intelligence is drawn directly from validated law enforcement data sources.
- This creates a mission-critical capability for field operations, as technical barriers to accessing intelligence systems disappear and real time intelligence becomes available to every officer, not just analysts.
We are moving toward a future where every patrol officer has the equivalent of an expert intelligence analyst at their side, able to instantly access and analyse the full scope of available intelligence through natural conversation. This democratisation of intelligence access, while maintaining strict data validity, represents a fundamental shift in how front-line law enforcement can leverage collected intelligence.
Transforming complex analytics into actionable intelligence
Another area where intelligence analyst chatbot makes difference is the ability of LLMs to convert sophisticated graph analysis into clear, actionable intelligence reports. Where analysts previously had to manually sift through vast amounts of network analysis data, LLMs can now synthesise this information into comprehensive, easily digestible narratives that reveal deeper patterns and insights.
For example, when analysing co-offender networks, LLMs can automatically process and narratively describe:
- Community structure and evolution over time
- Temporal patterns of criminal activity (time of day, day of week, seasonal trends)
- Geographic patterns using familiar area descriptions (“downtown,” “eastern suburbs”)
- Key player identification based on centrality metrics and PageRank
- Cross-community interactions and their evolution
- Crime type patterns and their relationship to specific locations or times
- Generate code snippets to retrieve specific data
This capability also transforms how complex intelligence is presented. By combining graph analytics with LLM-powered narrative generation, the system can not only identify patterns in criminal networks, but also help analysts explain them in ways that directly support tactical and strategic decision-making. GraphAware Hume Maestro with the right workflow in place can generate reports and help analysts create dashboards containing the key information in a visualised, easily accessible way.
Security is the main priority
GraphAware Hume is designed with a tight focus on access control and security in mind and Maestro conforms to that vision.
Christophe Willemsen,
GraphAware CTO
Due to the sensitive nature of the work of law enforcement and intelligence agencies, the primary focus when designing Hume Maestro has been security. “GraphAware Hume is designed with a tight focus on access control and security in mind and Maestro conforms to that vision,” explains Christophe Willemsen, Chief Technology Officer at GraphAware. Even if using LLM to query data in the knowledge graph, the users will not be able to get to the data they shouldn’t see.
He adds that LLM-based systems introduce new security challenges, and the team takes every precaution to protect client data and mitigate the generation of incorrect information. The clients can connect to their existing LLM providers such as OpenAI or Azure or their own local models via Ollama.
At the same time, confidential data from the organisation will not be leaked through prompts to the online LLMs – a significant worry for agencies working with sensitive information. “We are employing identification and anonymisation techniques to comply with security requirements of our clients,” C. Willemsen explains. Maestro therefore securely combines the power of LLMs with the law enforcement’s need for security, reliability and transparency.
If you want to see Maestro in action, register for our webinar or request a demo.
GraphAware, founded in 2013 in London, has become a leader in connected data analytics using graph technology, assisting analysts and data scientists around the globe. Their innovation stems from the belief that the world is interconnected, and traditional relational databases can’t capture these connections. Iulian Timischi, an expert in intelligence analysis with experience in the Romanian military and NATO, introduced fusion analysis via GraphAware Hume at an Interpol webinar.
Click here to find out more about how we help law enforcement organisations.
