Updated

How should the Definition of Done change when developers use AI?

According to the Scrum Guide, the Definition of Done is a formal description of the state of the Increment when it meets the quality measures required for the product. My take: AI does not change what Done means. It raises the cost of proving it. Make sure people can inspect the work, and a human approves it.

What the Scrum Guide says

The Scrum Guide does not mention AI. It does say what the Definition of Done is, and who must follow it:

"The Definition of Done is a formal description of the state of the Increment when it meets the quality measures required for the product."

"The Definition of Done creates transparency by providing everyone a shared understanding of what work was completed as part of the Increment."

"The Developers are required to conform to the Definition of Done." Read the Definition of Done section of the Interactive Scrum Guide.

Go deeper

Why AI Is Breaking Your Definition of Done (Part 3) is the video behind this page: who decides Done, and three things to add to yours. Code Got Cheap. Accountability Didn’t. walks through a team that enforces its Definition of Done while the work happens. For where AI-written work piles up first, read Why AI Won’t Speed Up Your Scrum Team (The 1999 Trap). This page is the short answer. For the playbook behind Part 2, read What does Anthropic’s AI-native SDLC playbook mean for Scrum teams?.

What I see in practice

The Definition of Done is the first thing to break. In my video Why AI Won’t Speed Up Your Scrum Team (The 1999 Trap) I said that when a team writes code faster with AI, “Your work in progress explodes. Your Definition of Done collapses.” The code lands in review and testing, which still run at human speed. In Part 2 I added: “Nobody can inspect it, so nobody can adapt. The Definition of Done quietly dies.”

My answer in Part 2, Code Got Cheap. Accountability Didn’t., is to enforce the Definition of Done while the work happens, “not argued about at Sprint Review.” The fictional team I walked through writes the failing test first and commits it. A plain script, with no AI in it, blocks the AI from editing the tests. And I quoted the line from Anthropic’s August 2026 playbook that matters most to me: “the agent that wrote the code has no way to approve it.” A human approves. As I put it: “If it doesn’t meet the Definition of Done, it doesn’t exist.”

Be careful what you count, too. In Synthetic Speed I wrote: “Stop measuring how fast your teams generate drafts, and start measuring how quickly working software reaches real users.” A Definition of Done that counts code written or tickets closed rewards exactly that. And when something goes wrong, I said it in the video: “Accountability remains 100% human.”

Common mistakes

What to do this week

  1. Read your current Definition of Done with the Developers. Ask which items an AI-assisted change could meet without any person having understood the code.
  2. Write down who approves AI-assisted work. It is a human. Name who.
  3. Bring one proposed change to your next Retrospective and inspect it after a Sprint. AI for Scrum Teams lists adjusting the Definition of Done and adding a human review step among the adaptations to try.

FAQ

Who decides what goes into the Definition of Done?

The Scrum Guide says: "If the Definition of Done for an increment is part of the standards of the organization, all Scrum Teams must follow it as a minimum. If it is not an organizational standard, the Scrum Team must create a Definition of Done appropriate for the product." My take: if rules about AI use already sit above your team, bring them into the conversation before the Developers change anything.

Can the AI check its own work?

In Code Got Cheap. Accountability Didn't. I asked the same thing: "If the AI writes the code and the tests, isn't that a student grading their own exam?" The setup I walked through has the tests written first, a plain script that keeps the AI from editing them, and a human who approves.

What if several Scrum Teams work on one product, and only one uses AI?

The Scrum Guide says: "If there are multiple Scrum Teams working together on a product, they must mutually define and comply with the same Definition of Done." My take: that means one bar for the product, whoever (or whatever) wrote the code. See also Will AI make my Scrum team deliver faster?

Related

Michael Vizdos

Written by Michael Vizdos

Certified Scrum Trainer® since 2006. About me · LinkedIn

35+ years working with teams around the world on Implementing Scrum. Found this useful? Feel free to share it with your team, or send feedback. I read all of it.

I sometimes use AI tools to brainstorm, draft and create, including images and video visuals. How I use AI.