Working Agreements
Working Agreements is a short activity for opening a retrospective. It usually takes between 10 and 30 minutes, depending on the size of the team. The goal is to establish a set of behaviors and attitudes that define the team and that everyone will work by from now on.
The team works together to generate ideas about effective working behaviors. They then choose three to seven agreements that will form the foundation of how the team works.
How to run the activity
The most common version takes between 10 and 30 minutes:
- Form pairs or small groups of no more than four people.
- Each group comes up with three to five working agreements that would help the team collaborate better. These should be new proposals, not agreements that already exist.
- Each group picks one of its agreements to present to the whole team.
- All the proposals are collected, and the team chooses between three and seven final agreements by vote.
If fewer than three agreements are proposed
You can use a thumb vote: thumbs up (agree), thumb sideways (I'll go with the majority), and thumbs down (disagree). Before voting, it's worth clarifying what each proposal means to avoid misunderstandings.
If more than seven agreements are proposed
You can use dot voting: each person gets three dots to distribute freely among the proposals. The result can be confirmed with a final thumb vote.
Examples of common agreements
- Start and end meetings on time.
- Listen without interrupting; anyone who wants to speak raises a hand or unmutes.
- Keep cameras on during synchronous meetings.
- What's said in the retrospective stays in the retrospective.
- If something isn't working, we say so in the retro, not in the hallway.
- Give and receive feedback with constructive intent.
Working Agreements in teams using AI
Bringing generative AI tools into everyday work—code assistants, content generators, chatbots—has made it necessary to set explicit agreements about how they're used. Without them, differences in how individuals use AI can create imbalances in workload, quality problems, or conflicts over authorship and accountability.
Some agreements teams are adopting:
- Transparency about AI use: if a member uses AI to generate code, documentation, or analysis, they declare it before peer review.
- Mandatory review: no AI-generated output is delivered without human review. Accountability for the result still lies with whoever signs off on it.
- Approved tools: the team agrees which tools may be used with project data, to prevent leaks of confidential or client information.
- Skill development: the team agrees which tasks use AI as support and which it prefers to handle without assistance, so as not to erode key competencies.
- Regular review: AI agreements are reviewed more often than the rest, given how fast these tools change.
Common mistake
Creating agreements and never revisiting them. A working agreement that no one remembers, or that's never updated, stops serving its purpose. The common practice is to read them at the start of each retrospective and revise them when the team feels they no longer reflect how it actually works. Without regular review, they become decoration.
Resources
📄 Notes: Agile RetrospectivesScrum Manager Blog (in Spanish) · Sep 2022
References
- Adkins, Lyssa. (2010). Coaching Agile Teams. Addison-Wesley.
- Derby, Esther; Larsen, Diana. (2006). Agile Retrospectives: Making Good Teams Great. Pragmatic Bookshelf.
See also
Want to take your agility further? You can look for upcoming course and exam dates, or go at your own pace by becoming a Agile Club member. Membership includes exclusive resources, e-learning classrooms, and access to Skill Arena: a space to practice and measure your agile skills whenever it suits you.