North gains structured private-market context

Cohere has launched a Crunchbase connector for North, its enterprise agent platform, allowing organisations to combine private-market data with their internal knowledge and workflows. The integration provides access to more than five million company profiles, funding and acquisition histories, competitor information and predictive signals. It is available to North customers that already hold an active Crunchbase licence and use single sign-on. Administrators enable the connector through organisational settings rather than giving every user a separate unmanaged data path.

The connector targets work that analysts and commercial teams often perform manually across databases, spreadsheets and internal systems. An investor could monitor portfolio risks, a strategy team could map an emerging market and a sales team could identify companies showing growth or funding signals. North can combine the external data with internal notes and generate a memo, target list or ongoing automation. The value lies in connecting research to a maintained business process rather than producing a one-off answer.

Crunchbase includes observed facts such as funding rounds as well as predictive outputs covering growth, risk, acquisitions and possible public offerings. Cohere says North identifies predictive figures as forecasts and distinguishes estimated valuations from reported values. That separation is essential. A model can make a forecast sound as definite as a filed transaction unless the interface preserves provenance. Users should be able to inspect the underlying record, date and methodology before relying on a signal for investment or outreach.

Agents can move from research to workflow

Cohere also says figures arrive with citations for human verification. Citations can reduce the time needed to audit an answer, but only if they point to the exact record supporting the claim. A generic company page or stale snapshot is insufficient for time-sensitive market analysis. Organisations should test how the connector handles contradictory sources, revised funding amounts, duplicate company identities and events announced after the underlying Crunchbase update cycle.

The bring-your-own-licence model means the connector respects the entitlements already established in Crunchbase and North. That avoids presenting the integration as a new unrestricted data licence. Administrators still need to map user roles between the services and verify that an agent cannot expose premium or restricted fields to colleagues who lack access. Generated documents and downstream automations may persist data after a user’s entitlement changes, so retention and revocation require explicit policy.

Citations distinguish facts from predictions

Market intelligence is particularly vulnerable to automation bias. A ranked list of acquisition targets or sales prospects can appear objective even when it reflects incomplete records, survivorship bias or model assumptions. North should support explanations for why a company was selected and what evidence would change the ranking. Teams should compare agent recommendations with historical outcomes and review whether certain regions, private companies or industries are systematically underrepresented in the source data.

Connecting Crunchbase to internal CRM or collaboration tools can create useful loops. A funding event might trigger an account review, while an acquisition signal could update a risk memo. Automation should not immediately send outreach or alter a forecast without controls. Triggers need freshness thresholds, deduplication and approval for customer-facing actions. A revised or retracted external record should also propagate to derived artefacts rather than leaving an obsolete alert embedded in an executive document.

Licensing and identity remain prerequisites

For Cohere, the launch strengthens North’s position as a governed enterprise workspace where external and internal context can be combined. Connectors are a competitive layer for enterprise agents because model quality alone does not determine usefulness. Buyers should compare coverage, permission fidelity, citation quality and the effort required to build reliable automations. They should also understand whether the connector is available in their deployment region and how its data crosses sovereign or private-cloud boundaries.

The Crunchbase integration is a practical addition for organisations that already license both services. It can shorten market research and keep recurring analyses current, provided forecasts remain visibly different from facts and entitlements survive every downstream step. The strongest implementation will use the agent to assemble evidence and maintain workflows while leaving high-stakes commercial decisions to accountable people. Speed is valuable, but traceable data and clear uncertainty are what make private-market intelligence trustworthy.

Security review should consider inference across datasets, not only direct field access. An agent may combine permitted Crunchbase facts with internal notes to reveal a confidential acquisition strategy or identify a company that a user could not find through either system alone. North policies should restrict sensitive automations and label outputs that merge external predictions with proprietary context. Teams also need strong controls over exporting generated target lists to email, CRM or shared drives. The connector is most useful when it brings evidence together, but that same combination can amplify commercial sensitivity and make a seemingly ordinary research request worthy of substantially higher scrutiny during routine operational use.