Palantir's AIP Analyst Adds Skills: Enterprise AI Begins Reusing Analysis Methods

Palantir Updates AIP Analyst with Skills and Analysis Lookup

On August 18, 2026, Palantir released an update to AIP Analyst, its natural-language analysis interface built on the Ontology platform. The update adds AIP Skills, Analysis Lookup, time-series analysis, Ontology interfaces, and the ability to export analysis results to Foundry resources such as Notepad, Quiver, Contour, and AIP Skill. The most significant additions are Skills and Analysis Lookup, which together introduce a mechanism for reusing analytical methods rather than merely retrieving historical content.

AIP Analyst: Tool-Calling Analysis Over Ontology

AIP Analyst allows users to ask questions in natural language. The agent can search Object Types and objects, build Object Sets, execute aggregations and Ontology SQL, and generate summaries, charts, or maps. The new version also supports time-series analysis and Ontology interfaces, expanding the range of data and business objects it can handle. Crucially, AIP Analyst preserves the entire analysis process—including enterprise data, Ontology resources, and tool invocations—and provides a Graph view to inspect each step's provenance, logic, and data transformations.

Skills: Packaging Reusable Analysis Methods as Callable Instructions

Palantir defines an AIP Skill as a set of "reusable instructions." Users manage available skills in the AIP Analyst Skills settings; when a skill is relevant to the current question, the agent loads it. Skills can be created or updated from the current analysis, and conversations can be exported directly as AIP Skills. The article emphasizes that a Skill represents a reusable analytical method, not a full snapshot of a single analysis result. For recurring problems, users can distill stable processing steps into a Skill so the agent can reuse them without re-specifying the same instructions each time.

Analysis Lookup: Historical Analyses as Templates, Not Live Results

Analysis Lookup lets AIP Analyst load a past analysis by its RID (Resource ID), showing which resources and tools were used previously. The loaded analysis serves only as a template: it does not carry the live results from that earlier run. If the current task requires those tools, they are re-executed against current data to produce fresh results. This design explicitly avoids the risk of stale historical values being mistaken for current state.

Method Memory: A New Layer Alongside Knowledge Memory

In typical agent designs, memory first addresses "what was said before" or "what documents contain"—storing chat history, documents, or other information for retrieval into context. This "knowledge memory" (the article's editorial term) remains important and is not replaced by Skills; AIP Analyst continues to support semantic search. Skills and Analysis Lookup introduce a complementary "method memory" (also an editorial term): for a class of problems, which resources and tools were used, and what analytical approach was formed. The article notes that Palantir does not frame this as a new Agent Memory architecture nor as a replacement for RAG or vector databases.

Deep Integration with Ontology

The feasibility of this mechanism rests on AIP Analyst being built atop Palantir's Ontology. Palantir describes Ontology as its architectural core, organizing data, logic, actions, and security into a unified business representation. AIP Analyst operates through a toolset that queries objects, runs aggregations, and performs other analytical operations on that representation. The official Ontology system architecture diagram illustrates this relationship.

Conclusion: From Accumulating Content to Accumulating Methods

Summarizing the update as "enterprise AI starts remembering how to analyze" requires precision: the system does not automatically ingest all human analytical experience. Instead, it provides two reusable mechanisms—Skills to save callable instructions, and Analysis Lookup to save the resource-and-tool structure of past analyses, re-executing on current data when needed. From a product-mechanism perspective, what enterprise AI can accumulate is extending from "past content" to "reusable analytical methods."

Sources (all Palantir official): Palantir Foundry Announcements: AIP Analyst now supports AIP skills, analysis lookup, time series analysis, Ontology interfaces, and Foundry exports (2026-08-18) – https://www.palantir.com/docs/foundry/announcements Palantir Docs: AIP Analyst - Capabilities – https://www.palantir.com/docs/foundry/aip-analyst/capabilities Palantir Docs: AIP Analyst - Using AIP Analyst – https://www.palantir.com/docs/foundry/aip-analyst/using-aip-analyst Palantir Docs: The Ontology system – https://www.palantir.com/docs/foundry/architecture-center/ontology-system