Pensieve

What is Pensieve?

Pensieve is a business-aware AI platform that builds a structured, living model of an organisation from the company’s existing tools and knowledge. Its core purpose is to help AI agents operate with full business context—so they can surface relevant information and learn from decisions rather than working from isolated fragments.

The platform ingests information scattered across documents, messages, calls, and code, then connects the dots between people, projects, decisions, and customers. It positions “business context” as the missing layer between raw data pulled on demand and the real understanding needed for decision-making.

Key Features

• Context Layer that ingests scattered company knowledge: Pulls from documents, messages, calls, and code to create a consolidated view of the organisation.
• AI-maintained knowledge graph: Builds a structured model of relationships between people, projects, decisions, and customers to support business-aware answers.
• “Surface what matters” agent behaviour: Applies frontier model intelligence to the full organisational context to proactively deliver what’s relevant before questions are asked.
• Conflict resolution and normalization: Normalises information and resolves conflicts between sources so the system behaves like a consistent source of truth rather than disconnected snippets.
• Outcome-aware learning loop: Tracks whether insights were actioned and what impact followed, using outcomes to improve over time.

How to Use Pensieve

• Join the waitlist to get access to the Pensieve platform.
• Connect your existing tools so Pensieve can ingest knowledge from the systems your team already uses.
• Use business context for AI agents: run agents that rely on the Pensieve knowledge graph to find patterns across communications and deliver relevant insights.
• Review outcomes over time so Pensieve can learn from whether insights led to action and what results followed.

Use Cases

• Compliance and audit readiness support: Identify gaps between planned work and commitments mentioned elsewhere (for example, a SOC 2 compliance promise referenced in sales discussions but not reflected in engineering planning).
• Revenue risk and pipeline monitoring: Surface early warning signals by linking sales narratives, pipeline details, and technical or operational progress to highlight deals at risk.
• Product and permissions planning: Detect recurring themes in customer feedback (such as increased demand for team-level permissions) and relate them to prior decisions and outcomes.
• Recurring decision tracking for teams: Help teams understand what’s driving repeated debates by remembering prior blocking factors and noting what has or hasn’t changed since the last decision.
• Decision memory for cross-functional discussions: Preserve the “why” behind decisions by connecting the decision itself with the surrounding context across projects and stakeholders.

FAQ

• What types of company information does Pensieve ingest? Pensieve is described as ingesting knowledge from documents, messages, calls, and code.

• How does Pensieve differ from tools that only pull data on demand? Pensieve focuses on normalization, conflict resolution, and building a structured, living source of truth, including decision memory and learning from outcomes.

• What is the “Context Layer”? The Context Layer is Pensieve’s complete picture of the organisation, built by converting scattered knowledge into a structured model (an AI-maintained knowledge graph).

• Does Pensieve learn from outcomes? Yes. The site describes an outcome loop where Pensieve sees whether insights were actioned and what impact occurred, improving over time.

• Is Pensieve a chat assistant or an agent platform? The content describes “deep work agents” that connect information across the business and surface insights using the knowledge graph, indicating an agent-based platform rather than a simple Q&A tool.

Alternatives

• General-purpose enterprise search + AI: Tools that index internal documents and let users search with AI can provide answers, but they typically focus on retrieval rather than maintaining a decision-aware, normalized knowledge graph.
• Data/BI platforms with dashboards: Business intelligence tools can analyze structured data, but they usually don’t automatically normalize conflicts across communications (messages/calls) into a living source of truth for AI agents.
• CRM or ticketing-centric AI: Systems that focus on one workspace (e.g., sales pipeline in CRM or tasks in ticketing tools) may help with isolated workflows, but they don’t inherently model relationships across projects and decisions from multiple knowledge sources.
• Agent frameworks with custom RAG: Developers can build business-aware agents using retrieval-augmented generation, but doing conflict resolution, decision memory, and continuous outcome learning requires additional integration and implementation work.