Course Overview

AI-Powered Backlog Management with GitHub Copilot, Azure DevOps MCP Server, and Agent Skills is a full-day instructor-led workshop for teams that want to move beyond ad-hoc AI prompting and apply AI as part of a repeatable, governed delivery workflow. Participants use custom agent skills to run structured backlog workflows: create hierarchical backlogs, refine work items, publish work items to Azure DevOps, and produce linked test cases from acceptance criteria.

Half the day is dedicated to hands-on labs in a realistic, isolated Azure DevOps environment. Participants build a real product backlog from a specification through to linked test cases, without manually creating, linking, and updating each work item in Azure DevOps. The emphasis is on building applied team workflows based on agent skills. Skills act as reusable and repeatable agent instructions that package standards, review steps, and approval gates, and guide the user through a sequence of proposals, manual reviews, and changes that automatically reach the delivery system only after explicit approval.

Course Description

AI-Powered Backlog Management with GitHub Copilot, Azure DevOps MCP Server, and Agent Skills is a highly practical workshop for product, engineering, and delivery teams that want to use AI inside real backlog operations, not only as an individual productivity aid. Participants work with GitHub Copilot, Azure DevOps MCP Server, and reusable agent skills to generate backlog structures, refine multiple work items, create linked test cases, and publish approved changes through human-controlled workflow gates.

Many teams already use AI agents for coding activities, while their backlog management often remains manual. Even when AI helps draft work items, someone still has to create them in Azure DevOps, link them to the right parents, refine acceptance criteria, update fields, check consistency, and keep related artifacts connected. Repeated across a sprint, that is not product ownership. It is operational drag. At the same time, many AI productivity approaches still rely on prompt templates: useful for one-off tasks, but fragile when a team needs repeatable results, shared standards, and controlled changes in a live delivery system.

A team delivery workflow should be properly designed as a sequence of controlled steps. It should know what happens first, what happens next, when to pause, what output to produce, and when a human must approve any change in a connected system. That is where custom agent skills matter.

In this workshop, you will experience using practical agent skill-based workflow patterns such as:

• Refining backlog content from a document, specification, or direct user input,
• Creating an initial hierarchical backlog and publishing a complete Epic -> Feature -> PBI structure as linked work items,
• Refining multiple work items locally before publishing approved changes to the server, and
• Generating linked test cases with useful test data based on work items and their acceptance criteria.

You will see how GitHub Copilot, enhanced by the Azure DevOps MCP Server, can work with a live Azure DevOps context, including work items, test plans, repositories, pipelines, and related delivery information. You will learn how to reduce manual backlog administration, improve consistency across work item refinement, preserve human review, and create a clearer handoff from product requirements to delivery-ready backlog artifacts. The agent skills used in the workshop show how teams can package standards, review steps, and approval gates into repeatable AI-assisted backlog and test management workflows.

Course Objectives

What You Will Learn

Participants work in a realistic Azure DevOps environment and learn how AI-assisted backlog and test management change when treated as a team workflow rather than an individual prompt. The instructor guides participants through the tool surface, the workflow model based on agent skills, and the human approval points that keep AI-assisted changes under control.

The practical work covers the full path from user requirements to delivery-ready backlog artifacts, connecting GitHub Copilot to the backlog environment through the Azure DevOps MCP Server, generating structured items from a specification, refining existing work items and acceptance criteria, and creating linked test cases that preserve traceability.

The agent skill patterns are shown at a practical level. Each workflow is built on a skill that packages the instructions, standards, and review steps for a specific backlog operation. For example: 

• Start from a document, specification, or direct user input and turn it into refined backlog content,
• Prepare and publish a full hierarchical backlog to Azure DevOps as linked work items,
• Locally refine multiple work items before publishing approved changes,
• Generate test cases and supporting test data from a selected work item and acceptance criteria, and publish them as linked Azure DevOps work items.

Participants leave with a practical understanding of how governed AI workflows can reduce manual backlog administration, improve consistency across work items, and create a clearer handoff from product requirements to delivery-ready backlog artifacts. They get their hands on a repeatable operating model their teams can adopt and adapt: AI helps produce content, people review the result, and approved changes are published only after the review is complete.

Who Should Attend

This workshop is designed for teams that already manage delivery in Azure DevOps and want AI to become part of a reliable workflow, not just an occasional writing aid. It is especially relevant for:

1. Product Owners and Business Analysts using Azure DevOps: For teams where backlog items are inconsistent, vague, too large, or not ready for implementation. The workshop shows how AI can support refinement, acceptance criteria, and clearer handoff to development.
2. Engineering Managers and Delivery Leads: For leaders whose teams lose time clarifying weak requirements before development can start. The workshop helps enhance backlog readiness and quality, enabling later AI-assisted implementation and review.
3. Scrum Masters and Agile Coaches: For people helping teams adopt AI responsibly, with guardrails, shared working agreements, and human-in-the-loop review patterns.
4. Product Ops and Transformation Teams: For organizations where AI adoption is fragmented across teams and needs a repeatable internal standard.

Learning Paths

Foundational AI-for-PM or AI-for-PO training is a useful starting point as it helps participants understand AI concepts and opportunities in product work. This workshop takes the next step: applying AI inside real team workflows, with backlog items, approval gates, and traceability.

If your team has already experimented with AI prompts, this workshop helps answer the next question: how do we make AI reliable enough for everyday requirements management work?

The course implementation is Azure DevOps and GitHub Copilot-specific, but the operating model is widely applicable: structured AI-assisted workflows, connected context, and human approval before changes are made. The same principles can be applied to tools such as Jira, ClickUp, or similar platforms that offer suitable AI integrations.

Prerequisites

No prior experience with agent skills or GitHub Copilot is required. Participants should have a basic understanding of how their team plans, refines, and tracks delivery work. Familiarity with Azure DevOps and knowing what work items, boards, and test plans are is sufficient. The lab environment is fully web-based and prepared for participants, so no local installation or configuration is needed. GitHub Copilot access is provided as part of the workshop environment

This is a hands-on course, so please have your own computer available.

How to Prepare

• New to GitHub Copilot? Spend a few minutes with the GitHub Copilot Quickstart or the OpenAI Prompt Engineering Guide. Both help build a basic feel for clear instructions and reliable AI outputs.
• New to Agent Skills? Review Skills in ChatGPT from the OpenAI Help Center. It explains how reusable skills help AI assistants follow specific workflows more consistently.
• New to MCP Servers? Review the Azure DevOps MCP Server Overview. It provides useful context for the workshop's technical foundation.
• If you are attending as a Product Owner, Business Analyst, Scrum Master, or Agile Coach, think of one recurring backlog or refinement activity you would like to make less manual.
• No software installation is needed. The lab environment is fully web-based and will be provided on the day.

Public Classes

Public workshops are open to the general public and run when a minimum attendance threshold is met. Because participants come from different organizations, industries, and roles, the discussion covers a wide range of Azure DevOps configurations, team structures, and governance scenarios. Attendees benefit from seeing how the same workflow patterns apply across different contexts.

Private Classes

Private delivery is the preferred option when an organization wants to align a team or multiple teams around the same AI-assisted backlog workflow. When participants work together regularly and attend together, they leave with shared vocabulary, shared governance expectations, and a workflow pattern they have already practiced.

The discussion and examples can be tailored to the organization's existing Azure DevOps project structure, naming conventions, and approval processes. Private classes are also useful when leadership wants teams to agree on practical AI working agreements before scaling adoption.

What Is Included

Why is this Approach Relevant Now?

AI in software delivery is moving beyond isolated prompts toward reusable, governed workflows: AI that works with real context, follows a defined sequence, pauses for validation, and asks for human approval before changing connected systems. GitHub describes this shift through agentic primitives, context engineering, MCP tool composition, and validation gates.

That shift matters directly for backlog management. Backlog items increasingly become input for AI-assisted delivery, implementation planning, test generation, and pull request creation. Microsoft notes that vague or overly verbose work items can reduce Copilot’s effectiveness.

In other words, backlog quality is affecting AI development readiness.

This workshop helps your team turn backlog management into a repeatable, human-controlled AI-assisted workflow in a concrete Azure DevOps environment.