The power of a connected AI system
See how Azure unifies AI, data, business context, apps, and agents to run an intelligent system across the enterprise.
Find solutions for putting your ideas into action
Unify IT and developer workflows with Azure Copilot and GitHub Copilot.
Build intelligent apps at enterprise scale with the Azure AI portfolio.
Find the right databases for your needs—including relational, NoSQL, and caching.
Realize value from data and generate near-real-time insights with advanced cloud-scale analytics solutions.
Achieve high performance for even your most compute-intensive AI workloads, including deep learning.
Accelerate AI innovation by securely migrating and modernizing your IT estate.
Unify siloed teams, distributed sites, and sprawling systems.
Move your SQL Server databases to Azure with few to no application code changes.
Run and manage applications your way with your preferred Linux distribution on Azure.
Run your SAP apps and workloads on the preferred cloud platform for SAP.
Realize value from data and generate near-real-time insights with advanced cloud-scale analytics solutions.
Enable data, analytics, and AI use cases on an open data lake.
Find the right databases for your needs—including relational, NoSQL, and caching.
Build AI-powered, intelligent apps and enhance your critical solutions with generative AI.
Deliver enhanced application and data experiences even faster.
Use AI-powered tools, services, and agents from Azure and GitHub to deliver continuous innovation and value to your developer teams.
Unlock insights with AI-powered integrations, an Azure-native API management solution, enterprise-grade security, and deep Azure connectivity.
Build intelligent apps at enterprise scale with the Azure AI portfolio.
Achieve high performance for even your most compute-intensive AI workloads, including deep learning.
Confidently scale the next generation of safe, responsible AI applications.
Explore tools for bringing your vision to life
Build, evaluate, and deploy generative AI solutions and custom agents.
Unify your teams and data to accelerate AI innovation with a complete data platform.
Connect agents securely to your systems, data, and workflows so they reason and act with your business context.
Get a powerful, lightweight, free code editor with integrated tools to develop, debug, and deploy all from one place.
Build and scale apps with managed Kubernetes.
Increase software development velocity and inspire continuous innovation.
Build and scale mission-critical applications on a new PostgreSQL cloud database service.
Build, evaluate, and deploy generative AI solutions and custom agents.
Find the right model from exploration to deployment all in one place.
Securely design, deploy, and scale AI agents with ease.
Build high-impact AI apps with one centralized solution for agentic retrieval-augmented generation (RAG) workflows.
Build cutting-edge, market-ready AI applications with out-of-the-box and customizable APIs and models.
Use AI to monitor text and image content for safety.
Use an enterprise-grade AI service for the end-to-end machine learning lifecycle.
Plan smarter, collaborate better, and ship faster with a set of modern dev services.
Increase software development velocity and inspire continuous innovation.
Seamlessly build, scale, deliver, and deploy secure software with AI at the core of your DevOps.
Get a powerful, lightweight, free code editor with integrated tools to develop, debug, and deploy all from one place.
Run AI apps and microservices using serverless containers.
Deliver AI-ready APIs with built-in governance, security, analytics, and Azure scalability.
Build workflows faster with AI, native Azure integration, real-time insights, and enterprise-grade security.
Deploy VMs in seconds.
Build and scale apps with managed Kubernetes.
Provision VMs for Ubuntu, Red Hat, SUSE, and other popular distributions.
Get industry-leading performance for SQL.
Deploy scalable, protected virtualized infrastructure.
Trigger apps with serverless computing.
Expand apps to additional locations.
Unify your teams and data to accelerate AI innovation with a complete data platform.
Find the right databases for your needs—including relational, NoSQL, and caching.
Develop AI-powered apps and agents with a fully managed and serverless NoSQL vector database.
Migrate and modernize your SQL workloads.
Innovate with a fully managed, AI-ready PostgreSQL database.
Accelerate app performance with a fully managed, scalable, in-memory solution.
Enable data, analytics, and AI use cases on an open data lake.
Unify on-premises, hybrid, and cross-cloud infrastructure.
Migrate and modernize your SQL workloads.
Build apps across on-premises, cloud, and the edge.
Simplify security operations with intelligent security analytics.
Protect multicloud and hybrid environments.
Create secure, private clouds.
Share code, track work, and ship software with modern dev services.
What grounds an Azure agent in your business context?
Fabric, through the Microsoft IQ layers. The bottleneck on most agent projects is not the model. It is that every new agent starts from zero, relearning how the business works, where data lives, what rules to follow.
Microsoft IQ brings together four capabilities. Which one you reach for depends on the context your agent is missing.
Captures how work happens.
Models how the business operates.
Enables agents to discover and reuse knowledge.
Adds real-time global context from the web. New at Build 2026.
Fabric IQ gives you three layers of business context
| Unified data | OneLake unifies the organization’s data estate, spanning analytical and operational data into a single, accessible layer. |
|---|---|
| Business intelligence | Semantic models provide structured, governed representations of that data which organizations already rely on for trusted business metrics. |
| Operational intelligence | Ontologies capture operational context by defining business entities and their relationships so agents can reason in the language of the business. |
Fabric IQ is generally available. Ontologies in Fabric IQ are expected to be generally available in the coming months. Operations agents are generally available now. Microsoft Databases can be centrally managed through the Database Hub in Fabric, currently in private preview, then mirrored into OneLake so operational data lands on the same foundation as analytics.
The stack-level view of Microsoft IQ lives on Microsoft AI. For the agent surface inside Microsoft 365 rather than Azure, see Microsoft 365 agents.
What makes Fabric Data Warehouse fast enough for agents to query?
GPU acceleration, built into the warehouse rather than bolted beside it. Microsoft integrated NVIDIA accelerated computing into Fabric Data Warehouse as query acceleration, which works automatically inside Fabric without requiring any query rewrites.
Microsoft calls this the first fully managed data warehouse to offer GPU acceleration. The research behind it was recognized by ACM SIGMOD as the “Best Industry Paper of 2026.”
The performance figures, with what they were measured against
| Microsoft benchmark | In internal benchmarking conducted in May 2026, the GPU-accelerated Fabric Data Warehouse delivered up to 7x faster performance relative to three comparable external vendors for reporting and application workloads at 64-user concurrency. |
|---|---|
| UNC Health | “We’re seeing up to 5x improvement in our query speeds, which allows our teams to spend less time managing performance and more time delivering meaningful insights,” said Shaun McDonald, IT Manager. |
The 7x is internal benchmarking against three vendors that are not named, not a published third-party test. The 5x is one customer measuring its own queries.
Why a warehouse needed this
AI applications are redefining how a data warehouse needs to perform. As AI agents reason over enterprise data, analytics systems need low-latency performance for many simultaneous users. With NVIDIA accelerated computing and custom CUDA kernels built directly into Microsoft Fabric Data Warehouse, Microsoft is bringing the SQL workflows customers already use into the production AI era.
Query acceleration is not something you can switch on today. Microsoft says it will be available for an early access preview in the next few weeks. Which Fabric workloads each license reaches is on Power BI in Microsoft Fabric.
How do you run an agent on Azure Functions?
You define it in an .agent.md file. Markdown instructions plus metadata that declares the trigger and tools, then you deploy it exactly like any other Function. No framework to wire up, no hosting infrastructure to manage. The runtime is in preview.
Any Functions trigger can run an agent
What the agent gets without extra code
The operational model does not change
Flex Consumption for scale-to-zero and per-second billing, managed identity for auth, Application Insights for traces, azd for deployment.
Two platform changes matter when you pick a runtime. Go is now a first-class language on Flex Consumption, in preview, laid out as a standard Go module. Rolling updates went generally available on Flex Consumption, so a deploy drains live instances in batches every few seconds rather than restarting them all at once.
How does an agent workflow survive a restart halfway through?
Durable Task Scheduler in Azure Functions. Agent work is rarely one request in one response. Long-running steps need somewhere to keep state when the process holding it goes away.
What standardizing on it unified for Microsoft Copilot
Microsoft opens its Build 2026 Functions announcements with its own Copilot team as the case study. As Copilot scaled to support complex, long-running AI workflows, its engineering teams needed a more reliable and consistent orchestration model. Microsoft says standardizing on Durable Task Scheduler helped run hundreds of millions of executions weekly while improving resilience and delivery speed.
That figure is what a single workload does on the service. Microsoft publishes it as a Copilot engineering result, not as a throughput limit for your app.
Rolling updates cover Durable workloads as well as HTTP and non-HTTP, so the zero-downtime deploy story applies to orchestrations rather than only to request handlers. The agent definition itself stays in the .agent.md file.
Can you do RAG on Azure without building vector infrastructure?
Knowledge as a Service in Azure Logic Apps, in public preview. It stands in for the plumbing that retrieval-augmented generation normally makes you build before you get to the part you actually care about.
What you would otherwise stand up first
What Knowledge as a Service does instead
The rest of what Build 2026 added to Logic Apps
| Azure Logic Apps Automation | A new Logic Apps SKU. Managed environment with a built-in AI assistant that takes you from intent to implementation in natural language. Public preview. |
|---|---|
| Azure Logic Apps MCP Server | Exposes existing Logic Apps workflows as MCP-compatible tools that agents can discover and invoke directly. Generally available. |
| Foundry agents in workflows | Invoke Microsoft Foundry agents directly from a Logic Apps workflow. Public preview. |
| Knowledge as a Service | Document ingestion plus retrieval as built-in workflow capabilities. Public preview. |
| Codeful Workflows | Code-first workflows for Logic Apps Standard, written with the Logic Apps Standard SDK in familiar .NET patterns. Public preview. |
| Migration Agent | Assessments plus migration guidance for organizations moving off BizTalk Server or third-party integration platforms. Public preview. |
Anything marked preview can change before it reaches general availability.
If the workflow belongs to the business team rather than the platform team, the Microsoft Power Platform low-code path reaches the same connector library. Ask for a quote to have both costed together.
Frequently asked questions
Yes. Operations agents are generally available too. Ontologies in Fabric IQ are expected to be generally available in the coming months. Two Fabric IQ integrations are still in preview: ontologies as knowledge sources in Microsoft Foundry, plus Fabric IQ as a first-party MCP tool in Microsoft Agent 365.
Yes, in preview, on Flex Consumption. The project is a standard Go module, so go build, go test and go mod tidy work as they normally do. Triggers in preview are HTTP, Timer, Service Bus, Event Hubs, Event Grid, Cosmos DB plus Blob Storage. There is no function.json to keep in sync.
Rolling updates, now generally available in the Flex Consumption plan. It is a configuration change. Instead of forcefully restarting every instance on a code or configuration update, the platform drains batches of live instances every few seconds while scaling out the new version to meet demand. It covers HTTP, non-HTTP and Durable workloads.
1,400+. It is the same managed connector catalog that already sits behind Logic Apps plus Power Platform, now available in Functions as first-class triggers with typed SDKs for calling connector actions from your function body. Managed connectors in Functions are in preview.
Yes. The Azure Logic Apps MCP Server is generally available. It exposes existing workflows as MCP-compatible tools that agents discover and invoke directly, so the automation you already run becomes agent-callable without a custom API in front of it.
By mirroring. Microsoft Databases can be centrally managed through the Database Hub in Fabric, currently in private preview, then mirrored into OneLake. That puts operational data on the same foundation as analytical data, which is what Fabric needs before it can make the data trusted, contextual and ready for AI.
Flex Consumption is the plan behind the serverless agent runtime. Microsoft describes it as scale-to-zero with per-second billing, so there is nothing running to bill between invocations. Deployment is azd, the same flow as any other Function app. Containers are not there yet: bringing your own OS-level dependencies to Flex Consumption as a container image is expected in the next couple of months, keeping the event-driven scaling with pay-per-execution billing.
Microsoft names two stages. It automatically ingests and processes the content, then it generates embeddings. Retrieval after that is a built-in workflow capability rather than infrastructure you run. Parsing and chunking are not called out as stages the service performs: chunking strategies and embedding models sit in Microsoft’s list of what retrieval-augmented generation normally makes you build first, not in its list of what the service does. Treat the internals as unspecified until the preview documents them.
Yes, in private preview. On-demand Sandboxes for Durable Task Scheduler move individual orchestration steps to managed, isolated compute while the orchestrator stays exactly where it is. You declare which activities should run as serverless, point at a container image, then Durable Task Scheduler handles provisioning, scaling and teardown. Each execution runs in a clean microVM-backed sandbox with per-activity or per-invocation isolation, which suits native toolchains, CPU-heavy preprocessing, cross-runtime steps and sandboxed execution of customer plugins or LLM-generated code.
What we do alongside the licence
SoftSolutionWorks.com is backed by BlueAlly, an authorized Microsoft reseller. As well as supplying the licences, we can help you choose the plan, deploy it, and finance it.