Connect. Sources plugged in, permissions intact.
Before the brain can think, it has to see — without ever seeing more than it should. Connect is where your systems plug in and your permission model survives the trip.
STAGE 01 · SOURCES PLUG INWhat happens at Connect
Connectors land on the systems you already run — no rip-and-replace, no big-bang migration. A first engagement typically connects the two or three sources that matter for the chosen function: the claims database and the CRM for an insurance workflow; the repos and the ticket tracker for an engineering one.
Permissions survive ingestion. The brain inherits your RBAC rather than replacing it: a query from an underwriter resolves against exactly the records an underwriter may see. Data never gets flattened into an all-access pool — access is evaluated at question time, not at copy time.
Every connector is read-scoped by default. Write access — needed later for write-backs — is granted per-tool, per-workflow, with its own token scope, so a connector that reads your ERP cannot silently gain the right to change it.
Connector guarantees
RBAC applied end-to-end
Source-system roles map into the graph. The brain never answers with data the asker couldn't open themselves.
Scoped credentials
Each connector authenticates with its own vaulted credential — rotated, monitored, and revocable in one click.
Change-data capture
Sources stay live. Updates stream into the graph, so the brain reasons on the business as it is, not as it was at migration.
Lineage from record one
Every ingested fact remembers its source system, record ID and timestamp — the foundation of stage 03's traceable rationale.
On-prem & VPC options
Connectors run inside your boundary when the data can't leave it — standard for defence, healthcare and BFSI engagements.
Six source families
People, ERP, CRM, Repos, Docs, Email — each with a dedicated connector page describing its nouns and workflows.
Explore →Common questions at this stage
Do we have to migrate or duplicate our data?
No. Connectors index and reference; your systems of record remain the systems of record. The graph stores structure and pointers plus the minimal payloads each workflow needs, under your retention rules.
What about systems with no API?
File-drop, database-replica and RPA-bridge connectors cover legacy systems. The Legacy Modernization capability can also map the estate first — see Capabilities.
How long does a first connect take?
For standard sources, days not months — the connect stage of a first function typically completes inside weeks 3–6 of an engagement.
Pick your function. Own the intelligence behind it.
Discover one opportunity, engineer one capability, deliver one measurable outcome — then scale.