Identity Protection is a service that enables organizations to view the security posture of any account. Organizations can accomplish three key tasks:
Automate the detection and remediation of identity-based risks.
Investigate risks using data in the portal.
Export risk detection data to third-party utilities for further analysis.
Always remember that Microsoft Entra Identity Protection requires a Microsoft Entra ID Premium P2 license to operate. Licensing is covered in more detail in a later unit.
Identity Protection uses the knowledge Microsoft has gained from its position in organizations with Microsoft Entra ID, the consumer space with Microsoft Accounts, and in gaming with Xbox to protect your users. Microsoft analyzes 6.5 trillion signals per day to identify and protect customers from threats.
The signals generated by and fed to Identity Protection can be further fed into tools like Conditional Access to make access decisions or fed back to a security information and event management (SIEM) tool for further investigation based on your organization’s enforced policies.
During production, you want to optimize and deploy your flow. Finally, you want to monitor your flows to understand when improving your flows is necessary.
You can optimize your flow by using variants, you can deploy your flow to an endpoint, and you can monitor your flow by evaluating key metrics.
Explore variants
Prompt flow variants are versions of a tool node with distinct settings. Currently, variants are only supported in the LLM tool, where a variant can represent a different prompt content or connection setting. Variants allow users to customize their approach for specific tasks, like, summarizing news articles.
Some benefits of using variants are:
Enhance the quality of your LLM generation: Creating diverse variants of an LLM node helps find the best prompt and settings for high-quality content.
Save time and effort: Variants allow for easy management and comparison of different prompt versions, streamlining historical tracking and reducing the effort in prompt tuning.
Boost productivity: They simplify the optimization of LLM nodes, enabling quicker creation and management of variations, leading to better results in less time.
Facilitate easy comparison: Variants enable side-by-side result comparisons, aiding in choosing the most effective variant based on data-driven decisions.
Deploy your flow to an endpoint
When you’re satisfied with the performance of your flow, you can choose to deploy it to an online endpoint. Endpoints are URLs that you can call from any application. When you make an API call to an online endpoint, you can expect (almost) immediate response.
When you deploy your flow to an online endpoint, prompt flow generates a URL and key so you can safely integrate your flow with other applications or business processes. When you invoke the endpoint, a flow is run and the output is returned in real-time. As a result, deploying flows to endpoints can for example generate chat or agentic responses that you want to return in another application.
Monitor evaluation metrics
In prompt flow, monitoring evaluation metrics is key to understanding your LLM application’s performance, ensuring they meet real-world expectations and deliver accurate results.
To understand whether your application is meeting practical needs, you can collect end-user feedback and assess the application’s usefulness. Another approach to understanding whether your application is performing well, is by comparing LLM predictions with expected or ground truth responses to gauge accuracy and relevance. Evaluating the LLM’s predictions is crucial for keeping LLM applications reliable and effective.
Metrics
The key metrics used for monitoring evaluation in prompt flow each offer unique insight into the performance of LLMs:
Groundedness: Measures alignment of the LLM application’s output with the input source or database.
Relevance: Assesses how pertinent the LLM application’s output is to the given input.
Coherence: Evaluates the logical flow and readability of the LLM application’s text.
Fluency: Assesses the grammatical and linguistic accuracy of the LLM application’s output.
Similarity: Quantifies the contextual and semantic match between the LLM application’s output and the ground truth.
Metrics like groundedness, relevance, coherence, fluency, and similarity are key for quality assurance, ensuring that interactions with your LLM applications are accurate and effective. Whenever your LLM application doesn’t perform as expected, you need to revert back to experimentation to iteratively explore how to improve your flow.
When you create a Large Language Model (LLM) application with prompt flow, you first need to configure any necessary connections and runtimes.
Explore connections
Whenever you want your flow to connect to external data source, service, or API, you need your flow to be authorized to communicate with that external service. When you create a connection, you configure a secure link between prompt flow and external services, ensuring seamless and safe data communication.
Depending on the type of connection you create, the connection securely stores the endpoint, API key, or credentials necessary for prompt flow to communicate with the external service. Any necessary secrets aren’t exposed to users, but instead are stored in an Azure Key Vault.
By setting up connections, users can easily reuse external services necessary for tools in their flows.
To create a Large Language Model (LLM) application with prompt flow, you need to understand prompt flow’s core components.
Understand a flow
Prompt flow is a feature within Azure AI Foundry that allows you to author flows. Flows are executable workflows often consist of three parts:
Inputs: Represent data passed into the flow. Can be different data types like strings, integers, or boolean.
Nodes: Represent tools that perform data processing, task execution, or algorithmic operations.
Outputs: Represent the data produced by the flow.
Similar to a pipeline, a flow can consist of multiple nodes that can use the flow’s inputs or any output generated by another node. You can add a node to a flow by choosing one of the available types of tools.
Explore the tools available in prompt flow
Three common tools are:
LLM tool: Enables custom prompt creation utilizing Large Language Models.
Python tool: Allows the execution of custom Python scripts.
Prompt tool: Prepares prompts as strings for complex scenarios or integration with other tools.
Each tool is an executable unit with a specific function. You can use a tool to perform tasks like summarizing text, or making an API call. You can use multiple tools within one flow and use a tool multiple times.
Tip
If you’re looking for functionality that is not offered by the available tools, you can create your own custom tool.
Whenever you add a new node to your flow, adding a new tool, you can define the expected inputs and outputs. A node can use one of the whole flow’s inputs, or another node’s output, effectively linking nodes together.
By defining the inputs, connecting nodes, and defining the desired outputs, you can create a flow. Flows help you create LLM applications for various purposes.
Understand the types of flows
There are three different types of flows you can create with prompt flow:
Standard flow: Ideal for general LLM-based application development, offering a range of versatile tools.
Chat flow: Designed for conversational applications, with enhanced support for chat-related functionalities.
Evaluation flow: Focused on performance evaluation, allowing the analysis and improvement of models or applications through feedback on previous runs.
Now that you understand how a flow is structured and what you can use it for, let’s explore how you can create a flow.
Before understanding how to work with prompt flow, let’s explore the development lifecycle of a Large Language Model (LLM) application.
The lifecycle consists of the following stages:
Initialization: Define the use case and design the solution.
Experimentation: Develop a flow and test with a small dataset.
Evaluation and refinement: Assess the flow with a larger dataset.
Production: Deploy and monitor the flow and application.
During both evaluation and refinement, and production, you might find that your solution needs to be improved. You can revert back to experimentation during which you develop your flow continuously, until you’re satisfied with the results.
Let’s explore each of these phases in more detail.
Initialization
Imagine you want to design and develop an LLM application to classify news articles. Before you start creating anything, you need to define what categories you want as output. You need to understand what a typical news article looks like, how you present the article as input to your application, and how the application generates the desired output.
In other words, during initialization you:
Define the objective
Collect a sample dataset
Build a basic prompt
Design the flow
To design, develop, and test an LLM application, you need a sample dataset that serves as the input. A sample dataset is a small representative subset of the data you eventually expect to parse as input to your LLM application.
When collecting or creating the sample dataset, you should ensure diversity in the data to cover various scenarios and edge cases. You should also remove any privacy sensitive information from the dataset to avoid any vulnerabilities.
Experimentation
You collected a sample dataset of news articles, and decided on which categories you want the articles to be classified into. You designed a flow that takes a news article as input, and uses an LLM to classify the article. To test whether your flow generates the expected output, you run it against your sample dataset.
The experimentation phase is an iterative process during which you (1) run the flow against a sample dataset. You then (2) evaluate the prompt’s performance. If you’re (3) satisfied with the result, you can move on to evaluation and refinement. If you think there’s room for improvement, you can (4) modify the flow by changing the prompt or flow itself.
Evaluation and refinement
When you’re satisfied with the output of the flow that classifies news articles, based on the sample dataset, you can assess the flow’s performance against a larger dataset.
By testing the flow on a larger dataset, you can evaluate how well the LLM application generalizes to new data. During evaluation, you can identify potential bottlenecks or areas for optimization or refinement.
When you edit your flow, you should first run it against a smaller dataset before running it again against a larger dataset. Testing your flow with a smaller dataset allows you to more quickly respond to any issues.
Once your LLM application appears to be robust and reliable in handling various scenarios, you can decide to move the LLM application to production.
Production
Finally, your news article classification application is ready for production.
During production, you:
Optimize the flow that classifies incoming articles for efficiency and effectiveness.
Deploy your flow to an endpoint. When you call the endpoint, the flow is triggered to run and the desired output is generated.
Monitor the performance of your solution by collecting usage data and end-user feedback. By understanding how the application performs, you can improve the flow whenever necessary.
Explore the complete development lifecycle
Now that you understand each stage of the development lifecycle of an LLM application, you can explore the complete overview:
Now that you have access to Copilot, set up your account, and have the latest tips to engage with Copilot you’re ready to get started with Copilot. Before you begin, it’s good to have a general sense of the technology behind this AI companion.
GPTs, or Generative Pretrained Transformers, are AI models that advanced our understanding of AI. These models excel at interpreting and generating human language. The ‘G’ in GPT stands for “Generative,” emphasizing their ability to produce original content. The ‘P’ stands for “Pretraining,” referring to the extensive training these models undergo on diverse datasets before they can be fine-tuned for specific tasks. The ‘T’ stands for “Transformer,” denoting the underlying architecture that enables these models to handle large amounts of data efficiently and generate complex outputs.
Microsoft Copilot is powered by a variant of the GPT model, designed to assist users with a multitude of tasks. This advanced AI technology enhances Copilot’s ability to provide valuable assistance in various scenarios:
Design Tasks: Copilot can help create images and other visual content, making it a versatile tool for designers.
Travel Planning: It can offer destination ideas, find flight prices, and help plan vacations, ensuring you get the best travel experience.
Cooking Assistance: Whether you need dinner ideas or specific recipes, Copilot can generate suggestions based on your preferences.
Fitness Guidance: Copilot can generate personalized workout plans and provide fitness tips to help you stay healthy and active.
Microsoft Copilot
Powered by GPT: Microsoft Copilot is built on a variant of the GPT model, designed to assist users with a multitude of tasks. By leveraging the advanced capabilities of GPT, Copilot can provide valuable support in areas such as design, travel planning, cooking, and fitness
Integration: Copilot is integrated into various Microsoft products, including Microsoft 365, Bing, and Microsoft Edge, making it accessible across different platforms and enhancing user productivity
Enhanced Capabilities: With its foundation in GPT, Copilot can understand and generate human-like text, cite sources, create poems, generate songs, and use multiple languages and dialects
GPT’s enable Microsoft Copilot to be an AI companion, enhancing your productivity and helping you navigate through various tasks with ease.
You’ve successfully set up your account and navigated your way here. Now, you might be wondering, “How do I engage with Copilot effectively?” The video is designed to help you get started. You’ll explore how to craft effective prompts, provide examples to spark your creativity, and show you how to make the most out of your interactions with Copilot.
https://go.microsoft.com/fwlink/?linkid=2299614
Clarify Your Objectives: Clearly articulate what you need. Whether it’s drafting a report, compiling research, or writing an email, giving clear instructions helps Copilot assist you effectively.
Provide Background Information: Context is crucial. Include relevant details about your task, such as the purpose, audience, and any specific requirements you have in mind.
Specify Sources and Parameters: If you need information from certain documents or specific data points, mention these explicitly to guide Copilot’s search and synthesis process.
Define Your Preferred Style: Let Copilot know the tone and style you prefer for your content. This could be formal, casual, technical, or creative, depending on your needs.
Be Detailed in Your Requests: The more details you provide, the better Copilot can tailor its responses to meet your expectations. For example, instead of requesting a brief overview, ask for a comprehensive summary with actionable insights.
Use Clear and Polite Language: Communicating clearly and courteously helps Copilot understand your needs better and ensures a more accurate response.
Engage in an Ongoing Dialogue: Keep the conversation dynamic by asking follow-up questions and providing feedback. This allows Copilot to refine its responses and better align with your requirements.
Understand Copilot’s Capabilities and Limits: Familiarize yourself with what Copilot can and can’t do. This helps you set realistic expectations and make the most of your interactions.
Microsoft Copilot is a multilingual AI companion designed to break down language barriers and make access to information easier across the globe. It currently supports several languages. However, language support for Copilot doesn’t extend as far as Bing’s language support, yet.
To adjust language settings in Microsoft Copilot, select the circle icon at the top right corner. Select on “language” and select your preferred language.
Aside from languages, Copilot was also built with accessibility in mind. Copilot voice is a feature that brings ease and efficiency to your conversations. It’s perfect for those moments when typing isn’t the best option, or for users who find speech more accessible than text.
Visual search is another feature that empowers you to explore and understand the world around you. Whether you’re curious about a mysterious plant in your garden or the species of a bird perched outside your window, visual search is there to assist. With a simple image, Copilot becomes your guide, transforming images into knowledge.
These features help pave the way to a future where knowledge knows no bounds, and learning is within everyone’s reach.
Your experience with Microsoft Copilot isn’t a one-size-fits-all. Whether you’re using a personal account or a work or school account, Copilot adapts to provide an experience that’s right for your needs. It’s compatible with various operating systems and browsers. You can also access Copilot by downloading the Microsoft Copilot app to your phone. To access Copilot from your computer, open a browser and go to copilot.microsoft.com.
You can use a limited version of Copilot without a Microsoft account. However, logging into a Microsoft account using the Microsoft Edge browser provides the best experience. Having an account allows for extended conversations with Copilot. It also allows you to save your conversation history and perform research and summaries of webpages you’re viewing.
Ready to get started?
First, navigate Microsoft Copilot. To sign in, locate the “sign in” option at the top right corner by clicking on the round icon. Now, you see the option to sign-in. When signing in with a personal account, you have access to a free basic option. If you don’t have a personal account, you can create one for free to access Copilot.
Microsoft Copilot is your intelligent AI companion designed to assist you with your daily tasks and enhance your productivity. Whether you need to find the latest research on a specific topic, draft a professional email, or seek advice on various matters, Copilot is here to support you. With its advanced capabilities, Copilot can help you navigate complex information, provide insightful suggestions, and streamline your workflows, making your day-to-day activities smoother and more manageable. Whatever the task at hand, you can rely on Microsoft Copilot to be your trusted companion in achieving your goals.
As a recap, to get started in Copilot follow these steps:
Open your preferred web browser: Start by opening the web browser you usually use.
Navigate to the Copilot website: Type copilot.microsoft.com into the address bar and press “Enter.” If you’re a Microsoft 365 user, you can also use copilot.cloud.microsoft.com.
Access through Bing: Alternatively, you can go to bing.com and look for the Copilot icon near the search bar. Clicking on the icon directs you to Copilot.
Use Microsoft Edge: Open the Microsoft Edge browser and look for the Copilot icon at the top right corner of the window. Clicking on the icon grants you access to Copilot’s features within Microsoft Edge.
Explore Copilot features: Once you are in, you can start exploring Copilot’s features like asking questions, summarizing information, organizing tabs, and more.
For Copilot Pro users: If you have a Copilot Pro subscription, you can access more features in select Microsoft 365 applications and get faster responses during peak times.
For businesses: Copilot is also integrated into platforms like Dynamic 365, assisting with data analysis, app development, and workflow automation.
Now that you successfully accessed Copilot, your next step is to explore and navigate the platform effectively.
