[Music] 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. Let's take a look at it in action 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. If 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. In 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. L alums 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 operational team can monitor from their dedicated 
workspace and adjust as any surprises come up. They'll also need to stay in the loop for 
certain steps by policy, such as verifying 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/.