PRODUCT

Text Business Intelligence

BI for the 80% of your data that isn't in a table.

Contracts, claims, tickets, reports, internal correspondence, structured into a queryable graph and exposed through dashboards. Every chart traceable to the paragraph behind it.

The data your BI tool cannot see

Most enterprise BI dashboards are built on the same 20% of data, transactions, line items, time series, things that already live in tables. The remaining 80% is text. Contracts, tickets, claims, internal reports, regulatory correspondence, meeting notes. The information that drives the business is mostly written down and almost never reportable.

You can search this text. You can ask an LLM to summarise it. You cannot, with off-the-shelf tools, turn it into a dashboard with trend lines, drill-downs and anomaly detection, because text does not naturally have rows and columns.

Text Business Intelligence is the layer that gives you both: structured analytics over unstructured text, with every visualisation traceable back to its source.

What Text Business Intelligence is

A BI surface built on top of BCNN's knowledge graph and forensic extraction stack. The system reads your text, contracts, claims, tickets, reports, emails, and turns it into structured entities, attributes and events the graph can understand. Standard BI components, charts, dashboards, drill-downs, alerts, sit on top of that structure.

The novelty is not the dashboard. The novelty is that the dashboard is built on an evidence-backed extraction. Every count, percentage and trend line links back to the specific text fragments that produced it. A surprising number on a chart can always be opened and read.

How it works

Three steps from raw text to a working dashboard.

  • Ingest. PDFs, emails, Confluence pages, ticket exports, free-text fields inside structured data. Native handling, no pre-processing required.
  • Extract. Forensic extraction parses documents against your ontology: who is mentioned, what was promised, amounts, dates, categories, sentiment, red flags. Entities and events land in the knowledge graph.
  • Surface. A dashboard layer queries the graph and renders the analytics. Click any bar in any chart and the source paragraphs open underneath.

Anomaly detection runs continuously. When a new clause appears in a contract, when a ticket category surges, when sentiment in customer correspondence shifts, the dashboard flags it before anyone has to look.

Capabilities

Forensic extraction over your ontology. Entities, events, monetary values, dates, sentiments and categories, extracted with citations.

Multilingual and multi-format. PDFs, emails, scanned documents, Office files, ticket systems, wikis. Polish, English and other languages handled natively.

Interactive dashboards with citations. Standard BI surfaces, trends, distributions, cross-tabs, geo views, with one important difference: every cell is clickable and shows the underlying text.

Anomaly and drift detection. New clauses in contracts, shifts in claim language, emerging topics in support tickets, flagged automatically.

Event-stream and batch modes. Real-time analysis on incoming text or batch analysis on archives. Both produce the same graph.

Open and integrable. Output as a queryable graph, an API, or pushed into existing BI tools (Power BI, Tableau, Metabase, Superset).

Where it's used

In customer operations to turn tickets into a real diagnostic: what categories are growing, what wording predicts escalation, where the support docs are wrong. In media and publishing to track narrative consistency, character arcs and editorial decisions across thousands of pieces of content. In security and OSINT to chart the evolution of a disinformation campaign across text and time.

Why Text Business Intelligence

Most "AI for unstructured data" tools give you a chatbot. A chatbot is a single-question interface. It cannot show you what is happening across 50,000 documents at a glance.

Most enterprise BI tools give you a dashboard over rows. The richest information in your organisation does not live in rows.

Text Business Intelligence is the tool that closes the gap. It is built on the same forensic graph as Networks Notebook and GraphRAG, which means the analytical surface and the question-answering surface share a single source of truth. Numbers on the dashboard match answers from the assistant. There is one truth, with citations.

FAQ

Is this just an LLM summarising my text?

No. The system extracts structured entities and events into a graph. The dashboards query the graph. LLMs are used during extraction and natural-language exploration, but the analytics themselves are deterministic.

Can I plug it into Power BI / Tableau?

Yes. The graph and its derived tables are exposed via standard interfaces. You can use BCNN's native dashboards, or pipe the structured output into your existing BI stack.

What if the extraction is wrong?

Every chart cell is traceable. Reviewers can audit specific extractions and feed corrections back into the model. Quality is measurable and improves over time.

How fresh is the data?

Either real-time (event streams from ticketing or messaging systems) or batch (daily, hourly). Configurable per source.

How does it relate to The Hub and GraphRAG?

The Hub answers natural-language questions with citations. GraphRAG is the underlying retrieval engine. Text Business Intelligence is the analytical surface, dashboards and aggregates instead of single answers. All three sit on the same graph.

Show us your text

We will run Text Business Intelligence on a representative sample of your unstructured data and produce a working dashboard in three weeks.

Request a sample dashboard →

Ready to explore?

See your text turned into a dashboard