PRODUCT · FLAGSHIP

Networks Notebook

The Digital Murder Board.

A knowledge graph engine with forensic BI/AI built in. Hundreds of thousands of data points connected into a single navigable network. Every insight traceable to its exact source.

Relationship Schema
Project ID: 8821-X
Swipe
EMAIL_LOGTRANSACTIONS.CSVOFFSHORE_REGARCHIVE_FILESREL: J. SMITHINTEL_REP_01.PDFSUBJ: A. VKOVTX: $4.2M [SWIFT]SHELL_CORP_LTDFIND_KYC_FLAG
Identified Entity
LLM Query
Contextual Data

When data is messy and being wrong is not an option

Investigations, due diligence, fraud cases, complex compliance reviews. They all start the same way: a pile of contracts, emails, transaction logs, photos, court documents. Tens of thousands of pages, dozens of systems, weeks of manual work before the first real question can be asked.

Standard tools fail in three different ways. Spreadsheets cannot capture relationships. Search engines index the surface and miss the connections. LLMs hallucinate, and even when they are right, they cannot prove it. For high-stakes work, probably right is insufficient.

Networks Notebook is built for the cases where being wrong has legal, financial or operational consequences.

What Networks Notebook is

A knowledge graph engine with a built-in BI/AI layer. It takes unstructured, multi-source data and turns it into a navigable, evidence-grade analytical structure: the Digital Murder Board.

The product does three things end to end. It builds the structure: ingesting documents, tables, images and database exports, extracting entities and relationships into a knowledge graph that respects your ontology. It supports the analysis: hypothesis-driven queries, network analytics, pattern detection, multimodal search across the graph. And it preserves the evidence trail: every answer, every visualisation, every conclusion links back to the exact source paragraph, row or pixel it came from.

You don't get a report. You get the evidence trail.

How it works

Four steps from raw data to actionable intelligence.

  • Ingest. Raw data in any form: PDFs, photos, tables, transaction logs, audio, video. Unstructured and semi-structured formats handled natively.
  • Structure. Extraction, cleaning, tagging, ontology mapping. The output is a typed knowledge graph aligned with the business logic of the case.
  • Analyse. Hypothesis-driven graph queries. Network analytics: communities, paths, centrality, anomalies. Multimodal retrieval across text, image and structured data.
  • Deliver. Clean dataset, relationship graph, analysis report, technical documentation. Everything cited, everything reproducible.

Capabilities

Multimodal mapping. Documents, images, audio and video are mapped into the same graph. A scanned contract, a CCTV still and a database row can sit on the same node and be queried together.

Hypothesis-driven analysis. Frame a hypothesis ("these three entities are connected through a shadow company") and verify it against the full interconnected structure. The graph either supports it, refutes it, or surfaces the missing link.

Total context mapping. Hundreds of thousands of data points connected into a cohesive network. Pre-built network algorithms surface communities, money trails, anomalous flows and structural patterns invisible in tabular views.

Traceable logic. Every visualisation, statistic and answer is linked to its source. Click a node, get the document. Click a number, get the rows behind it.

Model-agnostic, vendor-agnostic. Open-source or commercial LLMs. On-premise, sovereign GPU or fully air-gapped. Your data, your infrastructure, your control.

Where it's used

In AML and fraud cases to make hidden topologies (money mules, VAT carousels) visible at a glance. In logistics to correlate GPS, scheduling and external marketplace data into evidence accepted in court. In media to map a thousand episodes of narrative continuity. In OSINT to map the infrastructure behind disinformation campaigns at the level of structure, not individual posts.

The same engine drives every one of these. The ontology changes; the engine does not.

Why Networks Notebook

It is forensic-grade by design. Networks Notebook analyses have been accepted as evidence in court. Every claim cited, every model decision auditable.

It is explainable. Every answer cites its source. There is no black box because there is nothing to black-box: the reasoning is a path through a graph the user can inspect.

It is air-gappable. Full offline deployment, no external network calls, open-source models. For environments where data cannot leave the building, this is a precondition, not a feature.

It is fast. The standard pilot delivers usable results in three weeks. Most clients see their first hidden connection in week two.

FAQ

Is Networks Notebook a chatbot or a BI tool?

Neither. It is a platform with both BI and AI capabilities on top of the knowledge graph.

Do you need clean data?

No. The product is designed for messy, multi-source, multi-format input. We take care of cleaning and structuring everything for you.

How safe is it?

However you want! We can run our systems on your premises, in your cloud or in a fully air-gapped environment.

How does it relate to GraphRAG and The Hub?

Networks Notebook is the analytical workspace. GraphRAG is the question-answering interface on top of the graph. The Hub is an enterprise knowledge base with CRM capabilities. They share the same core but are packaged for different use cases.

What does a pilot cost and how long does it take?

Usually 3-6 weeks, €10,000-€15,000, on a real subset of your data and hypotheses you want to test. Result: expert analysis and report with key insights and recommendations, a working demo, and the AI agent working with your data.

Start a pilot

We define the scope, you provide a representative data sample, we deliver results in 3-6 weeks.

Start a pilot →

Ready to explore?

Run a 3-week proof-of-value pilot