The vocabulary behind every SIGNLD decision.

A database stores rows. A dashboard queries those rows and draws a chart. Neither knows how anything connects.

A Knowledge Graph stores the relationships too. A customer is linked to their contracts, invoices, support tickets and account manager. Contracts roll up to revenue. Projects consume cost.

That is why it can answer questions a single table cannot. "Why did churn spike?" spans your CRM, billing and support at once.

SIGNLD builds the graph for you. Connect your systems and it finds the entities, resolves duplicates across tools, and keeps the map current as your business changes.

Customers, contracts, products, invoices, costs, people, risks. Each is an entity.

The same entity usually exists in several systems under several names. Acme Corp in Salesforce and Acme in your ERP are one entity, not two.

SIGNLD resolves those duplicates when the graph is built, so every figure counts the thing once.

Relations carry meaning: a contract belongs to a customer, an invoice settles a contract, a project delivers it.

Reasoning happens across relations. Follow enough of them and a margin drop leads back to a single supplier price change.

Relations are inferred from your data, not designed up front. No schema work is required.

A metric, a customer segment, a renewal window, a board question, or a specific risk.

Each Topic runs continuously against the Knowledge Graph and produces a Decision Brief whenever something material changes.

In the product UI, Decision Topics appear simply as Topics in the left navigation. Add, rename or retire them at any time.

A Brief holds the question, the finding in plain language, the evidence behind every figure, and a recommended move with projected impact.

It is generated on demand, or automatically when a Topic shifts.

The point is that it is short enough to read in a meeting and sourced enough to defend in one.

Confidence reflects coverage and freshness: how much of the relevant data was available, how recent it was, and how consistent the systems were with each other.

It is shown next to the finding, not hidden behind it. A high-impact answer with thin coverage is labelled as such.

Low confidence usually points at a gap in a connected system rather than a wrong answer.

Traceability shows the path from a raw row, document or API response to the final figure.

That makes any claim auditable. A CFO can click a number and see the invoice behind it.

It is also how the platform stays honest: if the source cannot be shown, the figure is not asserted.

Connectors authenticate with the narrowest scope the system allows, and that access is read-only.

Nothing about your schema, records or workflows changes in the source tool.

Setup is measured in minutes per system, not in migration projects.

MCP lets SIGNLD reach systems that have no native connector, using a common protocol instead of bespoke code.

Gmail and Slack are connected this way today, alongside native connectors such as QuickBooks and NetSuite.

The same read-only rules apply: MCP connections read context, they do not modify it.

SIGNLD runs model inference on a single-tenant AWS Bedrock instance dedicated to your workspace.

Your content is not used to train models, and it is not shared with other customers.

Encryption in transit and at rest applies throughout. The full detail lives on the security page.