Microsoft this week previewed a software development kit (SDK) for its platform-as-a-service (PaaS) environment for building artificial intelligence (AI) applications, along with an extension to Microsoft Copilot Studio that makes it possible to build artificial intelligence (AI) agents.
Announced at the Microsoft Ignite 2024 conference, the SDK provides access to 25 templates that make it simpler to build and customize AI applications using AI Azure Foundry.
Microsoft CEO Satya Nadella told conference attendees that AI Azure Foundry is designed to make it simpler to discover and invoke multiple models that are now being routinely used within applications. Azure AI Foundry makes it easier for developers to experiment in a way that enables them to find the right models for their applications. “No application uses one model,” says Nadella.
Additionally, Microsoft is previewing tighter integrations between Azure AI Foundry and Microsoft Copilot Studio. Agents built in Microsoft Copilot Studio will be able to access an Azure catalog of more than 1,800 AI models. Application developers can also now use Microsoft Copilot Studio to work with images and voice-enabled agents.
There is also an Azure AI Agent Service, in preview next month, that will enable developers to orchestrate agents across business processes.
Developers can also build AI agents using the Microsoft 365 Agents SDK to access, for example, the Copilot Trust Layer that Microsoft has previously created, and the Microsoft Prompt Gallery that makes it simpler to share prompts has been updated to add support for AI agents. Microsoft Copilot Studio is the industry’s first agentic AI, said Nadella.
Microsoft is also making available a private preview of a reporting tool for Azure AI Foundry that makes it simple to create impact assessments for AI apps by assembling project details such as model cards, model versions, content safety filter configurations and evaluation metrics. These reports can be exported to PDF or SPDX formats for audits.
Other capabilities Microsoft is making available to AI application developers include updates to Azure AI Search that makes use of upgraded models to improve query performance. Azure AI Search will also soon be integrated with GitHub models.
Mitch Ashley, vice president and practice lead for DevOps and application development at The Futurum Group, said Microsoft is making a huge step forward. Developers can start with a library of predefined agents that can operate across Microsoft Teams, Office 365 the web and messaging platforms. That approach makes it possible to create highly sophisticated AI agents by leveraging Microsoft’s open source Semantic Kernel development to integrate AI models into their C#, Python or Java code.
Microsoft is also previewing an update to Azure Container Apps that provides access to graphical processor units (GPUs) using a serverless computing framework and a preview of Fabric Database that adds a SQL database that can be automatically replicated to the Microsoft data lake.
In addition. Microsoft has inked alliances with Weights and Biases and Gretel and Scale AI to extend the core capabilities of the Microsoft AI platforms.
Finally. Microsoft is adding AI security posture management capabilities to help security teams discover and map generative AI models and technologies across Azure OpenAI Service, Azure Machine Learning and Amazon Bedrock. Organizations can leverage built-in security recommendations to strengthen their security posture, identify infrastructure-as-code (IaC) misconfigurations and vulnerabilities that might have found their way into code and runtimes. It also maps attack paths that might impact sensitive data being exposed to an AI model.
Microsoft claims to have already helped provide 23 million people with AI and digital skills. The challenge now is to provide those individuals with the tools and platforms needed to infuse AI into every application imaginable.
