Palantir AIP empowers organizations to use large language models and other cutting-edge AI safely and securely. The platform allows you to do three things. First, AIP lets you deploy large language models and any other AI within your private network, anchored in your private data.

It helps you create a full-fidelity, real-time representation of all the concepts, actions, and decisions within your business—every truck, every warehouse, every shipment, every delay—that can then be used by AI.

Second, it enables you to define the ground rules around which data, LLMs, and AI can and cannot see, and what they can and cannot do on behalf of humans.

You can do this based on data sensitivity, laws and regulations, and competencies of the model.

AIP provides the framework for AI to hand off tasks to other trusted task-specific models, as well as to trusted humans. Finally, with this great power comes great responsibility. AIP brings industry-leading guardrails to control, govern, and trust the AI-powered business of tomorrow.

AIP automatically captures a full digital footprint of all AI inputs, outputs, and actions, including which AI was used for which decision, which humans were in the loop, and how they align with specific business purposes.

In this demo, let's say you're a leader at a US-based manufacturing company with customers all over the country. You've just received an alert that a hurricane is forecasted to hit one of your major distribution centers in the South, and you'll need to shut it down.

You want to know whether you should accelerate, delay, or cancel orders to your customers and other remediations you can take.

Here in AIP Terminal, you can ask questions and collaborate with a large language model, or LLM. "To what extent will Hurricane Cynthia affect our distribution centers in Louisiana?

" as well as follow-ups, "How does this impact customer orders and associated revenue?" Under the hood, as you ask questions, the LLM is able to traverse a data foundation powered by real-time data feeds from all across your business.

Importantly, you can control which parts of the business the LLM has access to. For example, we can see here that although the data foundation contains employee-level PII data the LLM can't see or access that data.

This has been informative. You want to understand your options and take action. You can ask AIP to simulate a baseline scenario.

What would happen if we shut down the affected Distribution Center? It uses a transportation network optimizer model and shows you the impact on backlogs, revenue, and commitments and highlights the impact on your highest priority customers.

You can also see which specific orders can be seamlessly fulfilled by other distribution centers.

This is a good start, but you want to do more. You want to know if you should charter additional trucks to increase the volume of deliveries before and after the closure. What are some options for chartering additional transportation for order fulfillment and the impact?

AIP offers three simulated courses of action alongside estimated cost.

Option B, a 20-truck capacity increase, seems to have the best cost-benefit trade-off, and your highest priority orders will all be taken care of. After validating it further, you propose your plan for cross-functional sign-off for manufacturing, supply chain, and customer service stakeholders.

Let's look under the hood again. Many of these steps required a deep understanding of how your business works.

LLMs alone don't have the right context, and we definitely don't want the LLM to give answers not grounded in reality. AIP connects LLMs with trusted models across your business, including forecasting, optimization, machine learning, and standard procedures.

AIP then facilitates a secure handoff between these models as needed, which keeps each model in its own lane.

This lets you then use the LLM plus all of your other models in concert, safely, to enable reasoning through the different scenarios and courses of action.

All stakeholders have approved your plan, and you now want your operational team to get cracking. Much of the chartering and rerouting has already happened through automation as you authorized. The customer consent for early deliveries as they make decisions and updates.

Routing and optimization models rerun autonomously on their behalf within the limits around key KPIs.

As decisions are made, they are continually pushed to your operational systems, triggering real-world action. They're also logged for future review. Let's go under the hood one last time and look at what enables your organization to safely govern LLMs and other AI to assist you through scenarios such as this one.

Here in the AIP control panel, you can set the guardrails and limits per model, including which data objects they can see, which actions they can recommend or take on behalf of humans, which tasks and workflows they are trusted to drive, who can use them, how cost is attributed, and so on.

You and your organization have formed an optimal, multi-pronged plan of attack that are taking real-world action within minutes, all using natural language within AIP, without a single line of code.

Introducing Palantir AIP | Capabilities and Product Demo

Palantir builds software that solves problems. We work with organizations across industries to transform how they use data and technology.

Learn more at https://www.palantir.com/.

Introducing Palantir AIP | Capabilities and Product Demo

Introducing Palantir AIP | Capabilities and Product Demo

Scripts