Microsoft Power Automate homepage

Microsoft Power Automate

by Microsoft

Microsoft's low-code automation platform, extended with Copilot Studio AI agents

Product with AI agentsSupervised

Last reviewed 2026-06-18

Visit websiteDocsAlternatives

Microsoft Power Automate is the automation product in the Power Platform family, covering cloud flows (API-triggered workflow automation) and desktop flows (RPA driving legacy and UI apps). It launched in 2016 as Microsoft Flow, was renamed in 2019, and gained RPA via the 2020 Softomotive acquisition. Its agentic layer sits at the seam with Microsoft Copilot Studio. "Agent flows" become tools that an agent's generative orchestrator calls at runtime: the flows are deterministic (same input, same output) while the surrounding Copilot Studio agent reasons over an LLM and decides which tools to invoke. LLM reasoning, multi-agent orchestration, MCP support, and model selection live in Copilot Studio, while Power Automate supplies the deterministic execution and a large connector library.

What it can do

  • Orchestrate agent actions with generative AI (Copilot Studio)

    Supervised

    An agent's generative orchestrator reasons over an LLM and decides which tools and flows to call to complete a request, governed by admin-set model and capacity controls.

    source

  • Run agent flows as deterministic tools

    Supervised

    Power Automate flows are exposed as deterministic tools an agent calls at runtime; given the same input they produce the same output, keeping execution on rails.

    source

  • Automate UI and legacy apps via desktop flows (RPA)

    Autonomous

    Drives legacy and UI-only applications via attended (supervised) or unattended desktop flows.

    source

  • Author and operate automation in natural language (Copilot)

    Copilot

    Builds and edits flows from natural-language prompts, assisting a human author rather than acting independently.

    source

Strengths

  • +Deep native Microsoft 365, Dataverse, Teams, and SharePoint integration plus a huge connector library
  • +Clean determinism-plus-reasoning split: flows run on rails while the agent layer reasons
  • +Strong governance and genuine multi-model support (GPT default, Claude and others selectable)

Limitations

  • −Confusing licensing (per-user vs per-bot vs per-flow), with agent flows consuming separate Copilot Studio capacity
  • −Fragmented agentic surface spanning GA and preview features
  • −Heavy Microsoft lock-in and SaaS-only, with no self-hosting

Overview

Microsoft Power Automate is the Power Platform automation product, spanning cloud flows and desktop-flow RPA. Its agentic capabilities live at the seam with Microsoft Copilot Studio, where deterministic flows become tools that an LLM-reasoning agent calls.

What it does

Agent flows are deterministic automations exposed as callable tools; a Copilot Studio agent's generative orchestrator decides which to invoke. Desktop flows automate legacy and UI apps (attended or unattended), and Copilot lets users author automation in natural language. LLM reasoning and model selection sit in Copilot Studio.

Integrations & setup

Native Microsoft 365, Dataverse, Teams, SharePoint, and Outlook plus 1,000+ connectors. MCP reached GA in Copilot Studio in May 2025; Dataverse can act as an MCP server.

Pricing

Power Automate Premium starts at $15/user/mo; Process is $150/bot/mo for unattended RPA. Agent flows consume Copilot Studio capacity billed separately.

Best for / not for

Best for enterprises already standardized on Microsoft 365 and the Power Platform. Less suited to teams wanting self-hosting or simple, predictable per-action pricing.

Alternatives

UiPath and Automation Anywhere are the closest RPA-plus-agents competitors.

What people are saying

We aggregate real LinkedIn discussion into sentiment for the agents people search most. Microsoft Power Automate isn't tracked yet, want it added? Request tracking.

FAQ

Is Power Automate autonomous?+

Its flow and RPA engine is deterministic, and Copilot Studio agents can run long operations autonomously via generative orchestration, but deployments are admin-governed with explicit human-in-the-loop action types, so in practice it operates as a supervised agent.

What models does it use?+

OpenAI and Azure OpenAI GPT is the default, with selectable alternatives such as Anthropic Claude available in Copilot Studio subject to admin enablement.

Sources

Last reviewed 2026-06-18

Alternatives & related

UiPath
Orchestrate AI agents, RPA robots, and people on one governed automation platform
Automation Anywhere
Goal-driven AI agents that orchestrate RPA bots, APIs, and people
Microsoft Copilot
Microsoft's AI: an in-app copilot plus a Copilot Studio agent platform
Zapier Agents
AI teammates that do work across thousands of apps using Zap actions as tools