Power Pilot Questions

Introduction

This documentation was created to clarify the main doubts related to the Power Pilot functionality. Here, you will find detailed answers about what Power Pilot is, how it works, its costs, integrations, and other important issues.

EstThis documentation does not aim to help you configure Power Pilot, but rather to answer any possible questions you may have about this feature. If you want to configure Power Pilot, follow the configuration documentation. However, we recommend reading this documentation first, as it will help you better understand the process: How to Configure Power Pilot.

What is Power Pilot?

Power Pilot is a generative AI developed to help you create customized assistants that efficiently answer your business questions.

Prerequisites and Operation

To use Power Pilot, you need:

1. Contract an AI model: It can be from OpenAI or Azure OpenAI. The use of the model is charged per token.

2. Register the model in the Power admin portal: After contracting the model, it needs to be configured in the admin portal.

Main Questions about Power Pilot

How Does the Cost of Power Pilot Work?

The cost is calculated by tokens:

• Input Token: Questions asked to the assistant.

• Output Token: Responses generated by the assistant.

PorFor example, with the GPT-4o-mini model:

• Input Cost: $0.15 per million tokens.

• Output Cost: $0.60 per million tokens.

You will only be charged when you reach one million tokens, making the feature accessible and economical. Additionally, if there is no interaction with the assistant, there will be no cost. AI Model Prices: https://openai.com/api/pricing/

O What is a Token?

Os grandes modelos de linguagem processam texto usando tokens, que são sequências de caracteres Large language models process text using tokens, which are sequences of characters (such as words or parts of words). These models learn the relationships between tokens and generate responses based on the sequences. NOn the OpenAI website, you can find a demonstration of what tokens are and a comparison with characters: Tokenizer: https://platform.openai.com/tokenizer

Is There a Monthly Cost Forecast?

400 questions asked to the assistant resulted in a cost of less than $5. Additionally, in the Power Embedded admin portal, there is a specific audit for Power Pilot usage.

This feature is very useful when you need to track the number of tokens used, estimate costs, and get an overview of consumption. With this, you can have a precise estimate of the costs involved.

How Does the WhatsApp Integration Work?

The integration with WhatsApp requires:

• A contracted AI model.

• An assistant configured in the portal.

Costs:

• Standard Number: R$150/month. This is the Power Pilot number, where the client cannot customize the photo or some information on WhatsApp.

• Customized Number: R$200/month. In this case, the client can use their own number, customizing the photo and some information on WhatsApp. (This number cannot be used on a cell phone; it must be a number solely for AI.)

Note: You can use a single WhatsApp number to manage multiple assistants. When a user interacts with Power Pilot on WhatsApp, only the assistants for which they have permission will be displayed.

Is it possible to grant permissions to specific users or groups?

Yes, when creating an assistant, you can define:

• Who will have access (specific users or groups).

• Whether the functionality will be available on WhatsApp or not for certain users or groups.

Is Power Pilot Different from Copilot?

Yes, both are artificial intelligence tools, but they are distinct:

• Power Pilot was developed by the Power Embedded team with a focus on providing business answers at a low cost. You can use this feature even with the smallest capacity of Fabric, such as an F2.

• Copilot, from Microsoft, is available in Power BI for users with a plan starting from the F64 SKU, which has a high initial cost.

Can I Use Power Pilot if I Use the "Show Reports in the Application" Feature?

Yes, it is possible to use the Power Pilot feature. However, since Power Pilot is displayed in the report navigation bar, you need to configure the HideNavbar parameter of the Identity API as true (hideNavbar = TRUE) to enable this feature correctly.

I Don't Have a Paid Azure Subscription. Can I Use the Free One?

To use the Power Pilot feature, it is recommended to have a paid Azure subscription. The free subscription offered by Azure has a token limitation, usually less than 10K tokens per minute, which is insufficient for ideal use of this feature. Power Tuning, as a Microsoft partner, can help you create your Azure subscription. Our cloud team can configure this subscription for you, allowing you to contract the necessary AI resource to use Power Pilot from it. Becoming a Microsoft partner is free. Simply fill out the form at the link below, and our cloud team will contact you:

Become a Microsoft Partner

Do I Need to Train the Model?

No, it is not necessary to train the model to use Power Pilot. All prompts are already configured internally via code. Simply follow the instructions in our documentation to configure Power Pilot and start using it immediately. Power Pilot Documentation

Can I Use My GPT Pro Contracted on OpenAI with Power Pilot?

No, GPT Pro is GPT as a service. In the case of Power Pilot, it is necessary to use the API of an AI model. When you need an isolated or customized AI, you can contract the APIs, and the client has the freedom to choose the model they want to use, such as GPT-4, GPT-4O Mini, among others.