Introduction

Scenario: You are orienting to how the defense portfolio brings generative AI to mission decisions without breaking the rules that protect classified work. In this unit you build literacy in AIP — what its builder tools do and why every AI action stays inside the same governance as a human analyst.

This unit is conceptual literacy, not a build: you will read the official AIP documentation and produce short written deliverables, not author any AI yourself. AIP is Palantir's generative-AI platform. It connects AI to your mission data and operations on a Foundry enrollment, working through the Ontology you met earlier so that models reason over real object types, properties, and links rather than disconnected text. Where Foundry is the data operations platform and Apollo is the continuous-delivery and operations platform underneath, AIP is the layer that lets large language models and agents actually do governed work against mission data.

AIP exposes a set of named builder tools. AIP Logic lets builders compose AI-powered functions and decision logic against the Ontology; AIP Chatbot Studio is used to build conversational assistants grounded in mission data; and AIP Evals provides a way to test and measure the quality of those AI behaviors before they are trusted. AIP Analyst is the agentic assistant that helps users explore and act on Ontology data through natural language. You do not need to operate any of these tools in this orientation — your goal is to recognize each one and pair it with a plausible mission use so a teammate understands which tool fits which job.

The reason AIP matters for defense is the governance guarantee. AIP and Foundry operate together as part of a shared service mesh powered by Apollo, and every operation made by humans and by AI agents abides by the same rigorous role-, marking-, and purpose-based controls that are cataloged in audit logging. In practice that means an AI agent cannot see, summarize, or act on data the requesting user is not cleared for: the markings and CBAC categories from the previous unit, the role-based permissions, and the audit trail all still apply when the actor is a model. By the end of this unit you will write a short explainer that maps each AIP tool to one mission use and states this governance guarantee in your own words, then read-only inspect a platform-seeded demo function to see that a governed AIP capability is real and runnable.

Capability focus: The AIP builder tools; mapping tools to mission uses; the governance guarantee (permissions/markings/audit apply to humans and agents). · Artifact: A short explainer mapping each AIP tool to one mission use + the governance guarantee.

Key concepts

Companion video

1. 1

   Read the AIP overview and list the builder tools

   Read the official AIP overview and confirm you can name the builder tools. The overview presents AIP as the generative-AI platform that connects AI with your data and operations, built on a Foundry enrollment and on top of the Ontology. From it, list the builder tools by their current official names: AIP Logic (compose AI-powered functions and decision logic), AIP Chatbot Studio (build conversational assistants), and AIP Evals (test and measure AI quality); also note the agentic AIP Analyst. This step is self-attested literacy: you confirm you have read the page and can restate the tool names — there is no artifact to machine-verify here.

   not startedself-attested

   Mark complete

   Self-attested: you can name AIP Logic, AIP Chatbot Studio, AIP Evals, and AIP Analyst.

2. 2

   Map each AIP tool to a notional mission use

   Write a short explainer that pairs each AIP tool with one plausible, notional mission use, drawing only on what the docs describe rather than inventing product behavior. For example, AIP Logic could power a structured 'recommend a course of action' function over mission object types; AIP Chatbot Studio could back a grounded assistant that answers analysts' questions from the Ontology; AIP Evals could measure whether that assistant's answers are accurate before it is fielded; and AIP Analyst could let a user explore Ontology data conversationally. Keep the mappings illustrative and labeled as notional. This is a self-attested deliverable: a human reviewer checks that each tool is paired with a sensible use — no read API confirms a written mapping.

   not startedself-attested

   Mark complete

   Self-attested: your explainer pairs each tool with one plausible mission use.

3. 3

   State the AI-governance guarantee

   State the AI-governance guarantee in your own words. AIP and Foundry operate together as part of a shared service mesh powered by Apollo, and every operation made by humans and by AI agents abides by the same rigorous role-, marking-, and purpose-based controls, all cataloged in audit logging. Explain concretely what this means for defense: an AI agent cannot see, summarize, or act on data the requesting user lacks the markings, CBAC categories, or role to access, and every AI action is audited just like a human's. This step is conceptual and self-attested — it is not API-confirmable; your deliverable is a clear written statement of the guarantee.

   not startedself-attested

   Mark complete

   Self-attested: you can explain that human and agent operations abide by role/marking permissions and are audited.

4. 4

   (Enrichment) Confirm a seeded demo AIP Logic function is executable

   Enrichment (read-only): inspect a platform-seeded demo published Logic function to see that a governed AIP capability is real and runnable — you are not building or authoring anything. The course environment provides a seeded demo published function named 'demoDisruptionRecommender'. Confirm it is enumerable and executable as a published query (the check lists the published query types and runs the seeded function read-only). Honest limits: this confirms only that a platform-seeded demo function exists and runs, not that you built it or understand its internals — there is no read API that introspects an AIP Chatbot's or Logic function's internal configuration, and AIP Agents v2 'Get Agent' is preview-only. Treat this as existence/run confirmation that AIP capabilities are governed and concrete, consistent with the literacy goal of the unit.

   not startedinstance check

   Run check

   Execute-don't-introspect: confirms a platform-seeded demo published Logic function is enumerable.