Scrumdemy
Practice examsCoursesCertificationsPricingBlog
Search exams and courses
EN
  • English
  • Español
  • Português
  • Français
  • Deutsch
  • Italiano
Sign inJoin free
Scrumdemy

Practice exams and courses for Scrum certifications. Independent; not affiliated with Scrum.org or Scrum Alliance.

Google PlayApp Store
Practice examsScrum MasterProduct OwnerScrum DeveloperAll exams
CoursesAll coursesTeach on Scrumdemy
ScrumdemyAboutBlogContact
FollowXLinkedInYouTubeInstagram

Scrumdemy

© 2026 ScrumdemyPrivacy · Terms · · English / Español / Português / Français / Deutsch / Italiano · Back to top
Scrumdemy
Practice examsCoursesCertificationsPricingBlog
Search exams and courses
EN
  • English
  • Español
  • Português
  • Français
  • Deutsch
  • Italiano
Sign inJoin free
Home/Blog

AI Agent Costs in Sprint Planning: Capacity Is Not Value

Nov 21, 2026 · 4 min read · Also in Español, Português, Français, Deutsch, Italiano

Many Scrum Teams now use AI agents that do more than autocomplete code. Agents write tests, migrate code, analyze logs and draft documentation, sometimes running for hours. That work is not free. Usage is usually billed by consumption, and a single ambitious task can cost more than expected.

This creates a new planning question: how should a team treat AI usage cost when it decides what to do in a Sprint? This article suggests a simple way to think about it using the 2020 Scrum Guide structure of Sprint Planning.

Sprint Planning in brief

Sprint Planning addresses three topics:

  • Why is this Sprint valuable? The Scrum Team defines a Sprint Goal.
  • What can be Done this Sprint? The Developers select items from the Product Backlog, considering their past performance, upcoming capacity and the Definition of Done.
  • How will the chosen work get done? The Developers plan the work needed to create an Increment that meets the Definition of Done.

AI cost belongs mostly to the second and third topics. It is part of capacity and of how the work is done. It is not part of why the Sprint is valuable.

Treat AI usage as capacity, not as value

Capacity is what the team can realistically spend in a Sprint: people's time, environments, budgets. AI usage is a new kind of capacity. It has a limit (a budget), it can run out, and it can be used well or badly.

The trap is to start treating it as a measure of progress. "We used a lot of agent time this Sprint" says nothing about value, just as "we worked a lot of hours" does not. A Sprint where agents produced a large amount of code that nobody needed is an expensive Sprint, not a productive one.

Making AI cost transparent

Transparency is one of Scrum's three pillars. If AI usage is invisible, nobody can inspect or adapt it. Some practical steps:

  • Know the budget. The team should know roughly how much AI usage is available per Sprint or per month, and who decides on changes.
  • Make big tasks visible. When a Sprint Backlog item depends on a long-running or costly agent task, note it in the plan.
  • Look at actual usage. Simple reports from the tools are enough. Review them at the Sprint Retrospective, not to blame anyone, but to learn.

Using cost in Sprint Planning decisions

When the Developers forecast what they can complete, AI usage cost can help them choose between options:

  • Several ways to do the same item. A large migration might be done with one long agent run, with small supervised runs, or partly by hand. The Developers can choose based on cost, risk and quality.
  • Exploration versus delivery. Letting agents explore many alternative solutions can be valuable but expensive. The team might set a limit for exploration within the Sprint.
  • Information for the Product Owner. If an item is far more expensive to deliver than it looked, the Product Owner should know, because it may change the order of the Product Backlog. Cost is part of value decisions, even if it is not value itself.

Who decides what

The Scrum accountabilities still apply:

  • The Developers own the Sprint Backlog and decide how to do the work, including which tools and how much AI to use within the agreed budget.
  • The Product Owner orders the Product Backlog to maximize value and benefits from knowing the real cost of items.
  • The Scrum Master helps the team and the organization make the cost visible and remove impediments, for example a budget process that is too slow or too rigid.

Budget limits themselves are usually set by the organization. If they block the team from reaching the Sprint Goal, that is an impediment to raise, not a reason to hide usage.

An example from Sprint Planning

A team plans to upgrade a framework across forty services. One Developer suggests letting an agent attempt all forty at once overnight. Another points out that a similar run last month was expensive and produced changes that took two days to review. The team decides to run the agent on five services first, review the results at the next Daily Scrum, and then continue in batches. The Product Owner hears the likely cost and agrees the upgrade is still worth it. Nothing dramatic, just transparent choices.

Quality still comes first

Cheaper is not always better. Cutting AI usage by skipping verification steps, or letting agents produce code without review, saves money in the Sprint and costs much more later. The Definition of Done applies to everything in the Increment, whoever or whatever produced it.

Signs your team handles this well

  • AI usage is discussed openly in planning and retrospectives, without drama.
  • Nobody uses agent activity as evidence of progress.
  • Expensive items trigger a conversation with the Product Owner, not a surprise at the end of the month.
  • The team improves its usage over time, for example by reusing prompts or breaking work into smaller, cheaper tasks.

Exam relevance

The Scrum Guide does not mention AI, so exam questions about AI focus on applying Scrum principles to new situations: who owns the Sprint Backlog, why transparency matters, and why accountability stays with people. PSM-AI Essentials covers how Scrum Masters can help teams use AI responsibly.

Prepare for AI-era Scrum questions

Try the free PSM-AI Essentials practice exam to test your thinking. For more practice, see the PSM-AI Essentials practice exams.

Practise for the exam

  • Free PSM-AI Essentials Practice Exam

More articles

Your first attempt can be your last.

Take a free 20-question exam now and see how far you are from the pass mark.

Start a free exam
Sign inJoin free