Advanced

Palantir AIP

Learn how Palantir's Artificial Intelligence Platform (AIP) integrates large language models with the Ontology for grounded, actionable enterprise AI.

What is AIP?

AIP (Artificial Intelligence Platform) is Palantir's approach to enterprise generative AI. It connects LLMs to the Foundry Ontology, giving AI models access to structured enterprise data, business context, and the ability to take governed actions.

Unlike standalone LLM applications, AIP grounds AI outputs in your actual data, reducing hallucinations and enabling real-world operational decisions.

💡
Key differentiator: AIP doesn't just answer questions — it can take actions. An AIP agent can analyze supply chain data, identify delays, and trigger rerouting actions, all within Foundry's security and governance framework.

AIP Components

💻

AIP Logic

Define AI workflows that combine LLM reasoning with Ontology queries, functions, and actions.

🛠

AIP Assist

Natural language interface for users to query Ontology objects and trigger actions through conversation.

📈

AIP Automate

Autonomous AI agents that monitor data, detect conditions, and take pre-approved actions automatically.

⚡

Model Integration

Use models from OpenAI, Anthropic, Meta, or custom models hosted within your security boundary.

How AIP Works

StepProcessExample
1. User queryNatural language input from user"Which shipments are at risk of delay?"
2. Ontology groundingAIP translates query to Ontology operationsQuery Shipment objects, filter by status and ETA
3. Data retrievalFetch relevant objects and propertiesReturn 15 at-risk shipments with details
4. LLM reasoningLLM analyzes data and generates responsePrioritized list with risk factors and recommendations
5. Action proposalSuggest Ontology actions to resolve"Reroute shipment S-1234 via alternate carrier"

AIP Security Model

AIP inherits Foundry's enterprise security:

  • Data access: LLMs can only access Ontology objects the user has permission to see
  • Action governance: AI-proposed actions require the same permissions as manual actions
  • Audit logging: Every AI interaction is logged with full context for compliance
  • Model hosting: Models can run within your security perimeter — no data sent to external APIs
  • Guardrails: Configurable safety controls on AI inputs and outputs
✅
Key takeaway: AIP's power comes from combining LLM reasoning with the Ontology's structured business context. This grounding approach reduces hallucinations, ensures governance, and enables AI to take real operational actions — not just generate text.