Overview

AIP Agent Studio allows users to build interactive assistants, known as AIP Agents, that are equipped with enterprise-specific information and tools, deployable internally in the platform and externally via OSDK and platform APIs.

AIP Agent Studio provides a natural language interface to leverage the Ontology, documents, and LLMs via AIP Agents that can take and update parameters (such as Ontology objects or text strings). For the following notional example, the following screenshot shows a LLM-powered AIP Agent that uses a parameter to take a filtered Ontology object set of customer support transcripts as context when answering user questions about current product issues.

The above AIP Agent can be deployed in a Workshop application.

AIP Agent Studio is built on the same rigorous security model that governs the rest of the Palantir platform. These platform security controls grant an LLM access only to what is necessary to complete a task.

Suitability

To understand whether AIP Agent Studio is the best Palantir platform tool for your workflow, consider the following questions:

• "How do I want to operationalize my workflow?", and
• "How do I use AI to automate workflows to gain insights from unstructured data?"

If your workflow would not require a "conversation-style" interface, we recommend you use AIP Logic or the AIP Generated Content widget in Workshop instead.

Examples where a conversation interface is not recommended include:

• A workflow with repetitive, well-defined tasks
• A workflow requiring speed and precision; conversations are an inherently open-ended method of communication and can lack the specificity of an interface where you set parameters directly

If a conversation interface is more applicable (usually for ad-hoc or human-empowering knowledge retrieval tasks), consider the following tiers of increasing complexity.

1. Ad-hoc retrieval-augmented generation (RAG)

   New to AIP or large language models (LLMs)? Start with AIP Threads to better understand how LLMs can help you improve productivity. Use AIP Threads for ad-hoc document analysis by dragging and dropping documents to get relevant LLM-powered answers.

2. Shareable RAG / basic agent

   Upgrade ad-hoc thread configurations from AIP Threads to AIP Agents for reusability, with granular permissions and configuration options in AIP Agent Studio. You can also create basic Ontology-powered, RAG-styled agents in AIP Agent Studio. You can use AIP Agents in AIP Threads or OSDK with platform APIs.

3. Part of a larger workflow

   Integrate AIP Agents into Workshop applications using AIP Interactive Widget or OSDK using Developer Console and platform APIs. Pass application-specific context to these applications with parameters such as object set variables.

4. Complex information retrieval and actions (soon)

   Define more complex or deterministic logic in AIP Logic or in Functions-on-Objects, and have the AIP Agent manage the conversation state. Use AIP Interactive Widget or OSDK with platform APIs.

Learn more about the core concepts of AIP Agent Studio or get started with building an AIP Agent.