AI assistant disclosure

The SalesSynq assistant uses an LLM provider to generate advisory text in response to your prompts. Outputs are advisory and may be inaccurate or incomplete. They never automate decisions affecting an individual's employment, credit, education, or essential services.

SalesSynq does not permit rep ranking or individual-performance use, including compensation, promotion, appraisal, discipline, PIP, dismissal or termination. Annex III employment and worker-management purposes and solely automated adverse decisions are prohibited.

What it does

• Summarises deals, signals, and pipeline data on request.
• Calls back-end tools to read or change data when you direct it to.
• Renders results inline as the same widgets you see on dashboards.
• Flags unresolved context gaps and, when strict same-company and same-product evidence is present, creates traceable deal attribution links for operator review and correction.
• Recommends next moves on deal pages. A user must choose Start before a task is created; SalesSynq does not autonomously update CRM fields.
• Accepts customer-approved, OAuth-authenticated MCP agent submissions into a deterministic GTM ingest ledger. Submitted payloads are not sent to an LLM by the ingest path.

What is logged

• Model version, hashed input, hashed output, and decision rationale per turn.
• Tool calls and tool results, with reason codes.
• Approval decisions for any tool that requires explicit user confirmation.
• Context-resolution decisions, including rule version, resolver version, candidate facts used, primary deal id, sorted attributed deal ids, actor, timestamp, action source, and evidence-manifest hashes.
• Recommended-next-move starts, including policy version, bundle version, context snapshot hash, feedback event ids, execution outbox id, SalesSynq task id, and optional Jira or Linear route reference.
• MCP GTM ingest events, including OAuth client id, agent user id, object type, idempotency key, payload hash, content hash, provenance, generated-content flag where supplied, materialisation status, quarantine reason, and replay count.

Conversation content is encrypted at rest in mcp_chat_messages; only hashes (not raw content) appear in the audit chain that backs the compliance dossier.

EU AI Act classification

SalesSynq publishes a preliminary intended-use classification and a versioned control catalogue. The classification is pending counsel review and is not a certification or a blanket claim of EU AI Act compliance.

Loading the current versioned control catalogue…

The machine-readable control catalogue records a preliminary limited-risk posture pending counsel review for advisory LLM outputs within the declared intended purpose. This is not a final legal classification or compliance claim. The full per-subsystem classification, model cards, and post-market monitoring record live in the compliance dossier exposed to administrators at /api/v3/compliance/dossier.

A preliminary EU AI Act classification record is published in a voluntary Annex IV-structured format. The record includes intended-purpose restrictions, classification rationale, human review controls, risk-management documentation, and post-market monitoring procedure.

Your rights

• You can opt out of AI-assisted suggestions per workspace.
• You can request export and deletion of conversation history.
• You can ask for a human reviewer where the assistant has produced an output you wish to dispute.
• You can correct context attribution in Pending Cases: human RESOLVE_GAP requires explicitly selected deal ids, and REJECT dismisses the unresolved gap.
• You can complete or dismiss follow-up tasks created from recommended next moves. For steps that require an external CRM update, the product asks for explicit acknowledgment before the task is started.
• Workspace administrators can revoke MCP OAuth clients and review or replay quarantined agent-originated GTM ingest events.