# BCNN

> BCNN is a European AI enablement company that prepares sensitive data for safe, explainable AI. Core offering: Data Forensics and Ontology as a Service - a 3-week process that turns unstructured documents, legacy databases, PDFs, and spreadsheets into a clean, structured dataset and a semantic ontology that your AI pipeline can actually use. Work runs on commercial cloud (AWS, Azure, GCP), sovereign European data centres, or fully on-premise air-gapped hardware. Outputs are forensic-grade and have been used as evidence in court. GDPR and EU AI Act compliant. Founders: Michał Domański (CEO) and Paweł Gołąb (CTO).

## Pages

- [Home](https://bcnn.io/): BCNN's approach to explainable AI and safe AI for organisations working with sensitive data. Covers the Networks Notebook semantic engine, the 3-week data preparation process, deployment scenarios (fraud detection, economic investigation, internal audit, employee fraud, corporate knowledge, insurance claims), and infrastructure options spanning commercial cloud, sovereign European GPU networks, and on-premise air-gapped hardware. Relevant for: explainable AI, safe AI, on-premise AI, sensitive data, European AI, data sovereignty, air-gapped AI, GDPR-compliant AI, EU AI Act.

- [Contact](https://bcnn.io/contact/): Get in touch with the BCNN team to schedule a discovery call or readiness assessment.

- [Home (Polish)](https://bcnn.io/pl/): Polish-language version of the BCNN home page, at bcnn.io/pl/. Same content as /, written for the Polish market: wykrywanie nadużyć, analityka śledcza, Networks Notebook, wdrożenie w chmurze, w suwerennym centrum danych lub on-premise. Relevant for: analityka śledcza, wykrywanie nadużyć, nadużycia pracownicze, AI on-premise, dane wrażliwe, RODO, AI Act.

- [Contact (Polish)](https://bcnn.io/pl/contact/): Polish-language contact page, at bcnn.io/pl/contact/. Form and booking link for starting an investigation with the BCNN team.

## Products

- [Networks Notebook](https://bcnn.io/products/networks-notebook/): BCNN's flagship product, a knowledge graph engine with forensic BI and AI built in. Described as the Digital Murder Board. Ingests PDFs, photos, tables, transaction logs, audio and video, extracts entities and relationships into a typed graph aligned with the client's ontology, then supports hypothesis-driven queries, network analytics (communities, paths, centrality, anomalies) and multimodal search. Every visualisation, statistic and answer links back to the exact source paragraph, row or pixel. Model-agnostic and vendor-agnostic; runs on-premise, on sovereign GPU or fully air-gapped. Analyses produced with it have been accepted as evidence in court. Standard pilot is 3-6 weeks at €10,000-€15,000. Used in fraud and internal audit, logistics, media and OSINT. Relevant for: knowledge graph, forensic analytics, link analysis, entity resolution, fraud detection, employee fraud, internal audit, procurement fraud, investigation software, multimodal analysis, court-admissible evidence, air-gapped analytics.

## Services

- [Services overview](https://bcnn.io/services/): Three services covering the range from data preparation to one-off forensic analysis and ongoing analytical capacity. Practitioner-led, outcome-scoped, evidence-backed. Relevant for: data services, analytics outsourcing, due diligence services, forensic analysis services.

- [Trash In, Tables Out](https://bcnn.io/services/data-preparation/): Trash in, tables out. Takes PDFs, reports, spreadsheets, CSVs, transaction records, internal documents, communications, public-source material, mixed-source case files and partially structured legacy exports, and returns clean structured tables, standardised datasets, resolved entity lists, linked relationships across sources, enriched metadata, source-linked records, data dictionaries and schema mapping. Five steps: ingest, clean, structure, link, deliver. Built for teams whose data is too messy or fragmented to use, organisations starting AI initiatives without usable source data, audit and compliance teams with fragmented records, and clients whose data cannot go through generic public-cloud pipelines. Works standalone or as the first layer before deeper BCNN analytics. Relevant for: data preparation, data cleaning, data structuring, entity resolution, ETL for unstructured data, AI-ready data, data quality, sensitive data handling.

- [Analytics as a Service](https://bcnn.io/services/analytics-as-a-service/): A senior analytical team (data engineers, knowledge graph specialists, domain analysts) engaged on a defined scope, working in the client's own environment on-premise, in their cloud, sovereign or air-gapped. Three engagement modes: single-question investigation (2-4 weeks, from €10,000) for board questions, regulator inquiries, M&A due diligence and internal investigations; recurring analysis cycles quoted per cycle for compliance reporting and executive dashboards; and embedded analytics where senior analysts work as an extension of the client team. Deliverables are running analyses rather than slide decks, every chart and conclusion links to the underlying data, and the client owns the reports, datasets, dashboards and code. Used in internal audit and employee-fraud investigations, in publishing and media for operational insight, and in public-sector institutions for grant and programme datasets. Relevant for: analytics outsourcing, data science as a service, embedded analytics, internal audit, employee fraud, procurement analytics, fraud analytics, forensic analysis, board reporting.

- [OSINT & Know Your Supplier](https://bcnn.io/services/know-your-supplier/): Forensic due diligence on suppliers, counterparties and business partners. Assembles evidence from EU and global registers (including KRS and REGON), sanctions and watchlists (EU, OFAC, UK, UN), litigation and insolvency records, beneficial-ownership databases, regulatory enforcement actions, news and OSINT, plus any internal data the client brings. Traces ownership networks across jurisdictions, including offshore and shell-heavy structures, and uses forensic entity resolution to connect the same entity across name variants, transliterations and registrations - the false negatives that defeat generic screening services. Where reputational or geopolitical risk is in scope, adds multimodal OSINT and disinformation footprint analysis. The deliverable is an executive summary with risk classification, detailed findings by risk category, a relationship graph, a sourced bibliography of every record consulted, and an explicit list of unresolved gaps. Available as an API call from a procurement or onboarding system, and extendable to ongoing monitoring. Relevant for: know your supplier, third-party due diligence, OSINT, beneficial ownership, sanctions screening, supply chain risk, enhanced due diligence, counterparty risk, disinformation analysis.

## Case studies

- [Case studies overview](https://bcnn.io/case-studies/): Five deployments across public sector, media, logistics and cybersecurity, filterable by industry and by the product or service used. Each is a real client, a real dataset, and a result that survived scrutiny - including in court. Relevant for: AI case studies, knowledge graph case studies, forensic analytics examples, public sector AI, on-premise AI deployments.

- [Logistics & Transport - Court-admissible evidence, from operational data.](https://bcnn.io/case-studies/logistics-fraud-detection/): A logistics company suspected systematic non-compete violations by drivers and subcontractors but could only produce suspicion, not evidence. BCNN fused GPS tracks, work-time records and listings from external transport marketplaces into a single behavioural model, and produced a court-admissible evidence report per driver and per subcontractor. The report, the evidence trail and the methodology were accepted by the court, and the client enforced its non-compete provisions on the strength of it. Built with Networks Notebook. Relevant for: fraud detection, logistics analytics, GPS data analysis, non-compete enforcement, court-admissible evidence, data fusion, behavioural analytics.

- [Public Sector - Experts free to do expert work.](https://bcnn.io/case-studies/cultural-institution-grants/): A grant-awarding institution faced large intake volumes against tight deadlines, with experts spending most of their time on formal checks instead of substantive review. GDPR and the data processing agreement ruled out cloud-based automation. BCNN deployed an on-premise hybrid system: deterministic rules for the unambiguous and quantitative checks, an AI expert layer for cases needing contextual interpretation, batches processed offline against anonymised personal data, and an optional KRS verification module. Verification time compressed dramatically and the risk of missing a formal error dropped close to zero. Built with Networks Notebook. Relevant for: grant review automation, public sector AI, GDPR-compliant automation, on-premise AI, document verification, hybrid rule-based AI.

- [Media - Perfect memory of a thousand episodes.](https://bcnn.io/case-studies/media-scenario-mapping/): A TV production company making long-running serial drama had outgrown manual archive search past a thousand episodes, and screenwriters were burning time verifying relationships, locations and past events. BCNN built a graph-based scenario-mapping system: every character, location and plot thread became a node, relationships were extracted from the script archive and kept current as new scripts arrived, and a natural-language interface let the team ask continuity questions directly. Fact-verification time dropped sharply with a direct effect on production costs, new writers came up to speed in days rather than weeks, and the graph became an asset for spin-offs and adjacent productions. Built with Networks Notebook. Relevant for: knowledge graph, media production, script analysis, continuity management, narrative graph, archive search, natural language querying.

- [Media - One platform to manage the entire publishing process.](https://bcnn.io/case-studies/publishing-process-optimization/): A publishing house with multiple imprints ran the entire book lifecycle on spreadsheets and email threads, leaving actual status invisible and letting a slip on one title cascade into others. BCNN shipped a dedicated web application, deployable in cloud or on-premise: a publishing calendar as the single source of truth, a status workflow on a visual timeline, and automatic notifications on schedule changes. Rollout was phased, starting with a quick automation layer over the existing Excel files that delivered value inside weeks. Manual status reporting was replaced by an always-current timeline, and the modular architecture is ready for automated proofreading and fact-checking modules via API. Built with Networks Notebook. Relevant for: publishing workflow, book production management, process automation, workflow software, editorial operations, project management app.

- [Cybersecurity - Not just the fake news. The whole machine.](https://bcnn.io/case-studies/osint-disinformation-monitoring/): A state institution needed to counter disinformation around politically sensitive events, where conventional media monitoring surfaced individual pieces of false content but missed the structure behind them. BCNN deployed a multimodal AI monitoring system that analyses text and image jointly, classifies content by topic, sentiment and propaganda markers, and uses graph analysis to identify coordination patterns: synchronised posting, shared content with subtle variation, network-amplification topologies and account-creation bursts. The output is a map of the infrastructure behind a campaign rather than a list of suspicious posts, allowing the client to disrupt the machine rather than chase individual posts. Built with Networks Notebook. Relevant for: OSINT, disinformation detection, botnet detection, troll farm analysis, social network analysis, multimodal AI, propaganda monitoring, information warfare.