Iteration done right: Agile & Scrum for Product Managers
Every product team you will ever join says it is "agile." A PM who really understands sprints, flow and honest estimation can tell the difference between a team that is learning fast and one that is just holding more meetings. In 2026 there is a new twist: AI agents now pick up tickets, open drafts and ask for review, right alongside people. The PM who can set clear goals, limit work in progress and define "done" for both humans and agents is the one who keeps the whole team moving. This course is about delivery. If you are looking for Agile as a way of learning, see our related reading, Agile Learning.
What you'll be able to do
Run Scrum end to end
Explain the three accountabilities, five events and three artifacts of Scrum, and facilitate sprint planning with a clear, testable sprint goal.
Manage flow with Kanban
Set work-in-progress limits and read flow metrics such as cycle time and throughput, so you can spot a stuck board before the deadline does.
Refine and estimate honestly
Keep a backlog healthy through regular refinement, write small user stories, and size them with story points without turning estimates into promises.
Read the charts
Use velocity, burndown and cycle time to start good conversations, and run retrospectives that end with one change the team actually tries.
Understand scaling basics
Describe, in plain English, how several teams share a roadmap without a giant process, and when a framework is worth the overhead.
Lead a human + agent team
Assign work to AI agents, write a definition of done for agent output, and keep a person accountable for every review.
Your certification path
Every step below is free, so the whole recommended path costs $0. You finish with a verifiable IBM SkillsBuild badge and a Microsoft Learn achievement, plus a free Scrum.org practice score that shows you are ready for the optional paid exam, if you ever want it. Hours are our own estimates.
Step 1 · Read the two official guides
Provider: Scrum.org / Scrum Guides (Scrum Guide) and Kanban Guides. You earn: the shared vocabulary every other step relies on (no certificate). Cost: Free · About 2 hours
Scrum Guide
Kanban Guide
Step 2 · Earn the IBM Agile Explorer badge
Provider: IBM SkillsBuild. You earn: the Agile Explorer digital badge (Credly). Complete three courses (A Taste of Agile, Agile Operations Fundamentals, Agile Program Fundamentals) and score 80% or higher on each end-of-course assessment. Register as an adult, student or educator on SkillsBuild first. Cost: Free badge · About 6 hours
See the badge requirements
Sign up at SkillsBuild
Step 3 · Microsoft Learn: Plan with DevOps
Provider: Microsoft Learn. You earn: a module achievement on your Microsoft Learn profile after you pass the knowledge check. A beginner module with 8 units, including "What is Agile?" and agile planning on boards. You do not need an Azure account for the reading units. Cost: Free course · About 1.5 hours
Open the module
Step 4 · Scrum.org Open Assessments (practice)
Provider: Scrum.org. You earn: a score to measure yourself, not a certificate. The Scrum Open assessment is free and can be retaken. Aim for 100% a few times in a row before you call yourself ready. Cost: Free practice · About 2 hours
Take the Scrum Open
Optional, later · Professional Scrum Master I (PSM I)
Provider: Scrum.org. You earn: the industry-recognized PSM I certificate. 80 questions, 60 minutes, 85% to pass. Cost: Paid exam, $200 USD per attempt. This is not part of the $0 path. Only consider it once you are consistently scoring well on the free Open Assessments, and check the current price on the official page first. About 20 hours of study
PSM I details (paid)
Practice in GitLab
Do these on your real project board in your own GitLab workspace. No code and no new projects needed. Our GitLab is Ultimate, so iterations, board WIP limits and scoped labels are all available to you. If a menu is missing, it is probably hidden by your role, so ask your coach.
1. Run one real iteration
Open Plan › Iterations in your group, create an iteration cadence of one or two weeks, and pull 5 to 8 issues into the current iteration. Write the sprint goal in one sentence at the top of the iteration description.
2. Build a sprint board with workflow labels
Create scoped labels workflow::ready, workflow::in progress, workflow::in review and workflow::done. Because they share the workflow:: prefix, an issue can only hold one at a time. Then make an issue board with one list per label, filtered to the current iteration.
3. Set WIP limits and estimate with weights
On the in progress list, set a work-in-progress limit (try 3 for a team of 2). Then give each sprint issue an issue weight using story points (1, 2, 3, 5, 8). Anything over 8 gets split into smaller issues. Agree as a team what happens when the limit is hit: finish something before you start something new.
4. Watch the burndown, then run a retro
Midway through the iteration, open the iteration page and read its burndown chart. Is the line flat? That is a conversation, not a verdict. On the last day, create a "Retro" issue with three headings (What went well, What was hard, One thing we will change) and a checklist. Next sprint, copy it with the /clone quick action instead of starting over. A maintainer can later turn it into an official issue template.
The agentic angle
Agents do not change what Scrum is for. They change who can hold a ticket. In a healthy agentic sprint, an AI agent is a team member with a clear job, a visible card on the board and a human who answers for the result.
- Agent-owned issues: mark them with a label such as
owner::agent and name the human reviewer in the description. Small, well-bounded issues (weight 1 to 3) suit agents best.
- A definition of done for agent work: the acceptance criteria are written, the output is linked in the issue, sources or assumptions are listed, and nothing moves to
workflow::done until a person has reviewed it.
- Human review is a WIP limit too: an agent can produce drafts faster than people can check them, so cap the in review list.
- Measure it like any other work: compare cycle time and rework for agent-owned and human-owned issues at the retro.
Exercise: sprint planning with an AI teammate
Paste your sprint goal and backlog titles into Claude (or another assistant) and ask it to: (1) flag issues that are too big and propose splits, (2) suggest story-point weights with its reasoning, and (3) draft acceptance criteria for two issues. Then decide for yourself: keep, change or reject each suggestion and write down why. Finish by writing a 4-line "definition of done for agent work" and paste it at the top of your board's description. Never paste confidential information into a public AI tool.
Free resource library
GitLab and Microsoft docs
Related reading on this site: Agile Learning, which looks at agile ideas from an education point of view.