dooopSoftware · Organization · 4 min
How to Coordinate AI Practices Across Different Squads
Define common criteria and local adaptations for AI among squads. Organize exceptions, evidence, and authority in shared decisions.
Published on September 6, 2026
CENTRAL THESIS
Coordination must distinguish necessary differences from lack of agreement.
A common minimum can preserve quality without imposing the same procedure on every task.
Coordinating AI practices across squads requires distinguishing what needs to be common and what can vary according to the task. A single rule for every situation may overlook important differences; the absence of shared criteria makes it difficult to review the whole. Start with decisions that cross teams, establish a common minimum, and define how local adaptations will be recorded, evaluated, and incorporated when appropriate.
Locate Shared Decisions
Look for points where one team depends on another's work: reused context, common components, integration of changes, review criteria, and responsibility for incidents. These points require understandable agreements between parties.
Do not start with the catalog of allowed tools. Tool choice may be a consequence of requirements but alone does not explain how the organization will maintain quality and continuity.
The DORA 2025 presentation describes AI as an amplifier of organizational strengths and weaknesses. The proposed application is to observe which differences between squads are necessary for the work and which represent a coordination gap.
Define a Common Minimum per Decision
The common minimum can state that every change must have intent, evidence of verification, and a responsible party for acceptance. The way to produce this evidence can vary according to the product and task.
A useful agreement answers:
- Which decision requires shared criteria.
- What evidence must exist in any team.
- What each squad can adapt.
- Who decides on an exception affecting other teams.
- How the organization monitors the effect of the adaptation.
Avoid using the same detailed procedure for tasks with different conditions without examining its function. Also avoid calling any local preference a necessary exception. The justification must rely on context and the behavior that needs to be preserved.
Make Local Adaptations Produce Reusable Information
A team may experiment with a different way to prepare context or review outputs. Record the problem, the change made, the criteria observed, and the limit of the conclusion.
Microsoft ExP describes the link between hypothesis, measurement, and iteration. In coordination between squads, use this discipline so that a local practice returns as examinable evidence, not just as a recommendation from someone who liked the tool.
An adaptation that works in one context may remain local. To turn it into a shared reference, examine whether other teams can reproduce the necessary conditions and which differences need to be preserved.
Fictional Example: Review in Products with Different Rhythms
Imagine an organization with one squad that changes internal interfaces and another that maintains shared integrations. Both use AI in preparing changes but require different checks before acceptance.
The common agreement requires intent, revised scope, and evidence of functioning. The first squad may use interaction cases and visual inspection. The second needs to examine contracts and effects on consuming teams. The difference in method does not eliminate the common commitment to evidence.
If an integration change affects the interface, the flow must indicate when both teams participate in the decision. The example shows how to combine shared criteria and local execution without assuming one procedure will suffice for all products.
Maintain a Path to Resolve Disagreements
When squads disagree, reconstruct the disputed decision. There may be a conflict of criteria, lack of information, or different legitimate needs. The meeting should end with a choice, a limited investigation, or a recorded exception.
Define who has authority when the decision goes beyond one team. Do not leave this role implicit in the seniority of the loudest speaker or the fact that a tool is already in use.
DORA on documentation highlights clarity, ease of location, and reliability. The shared agreement should point to the current version and explain its exceptions. Parallel documents without clear relation make it difficult to know which criteria apply to the task.
Review Coordination Without Erasing Local Needs
Monitor signs of problems in the whole: changes returned between teams, contradictory criteria, context that does not reach the destination, and decisions without responsibility. Choose interventions linked to these signs.
If the common rule prevents a legitimate task, revise the rule or record the exception with evidence. If the local adaptation transfers difficulty to another team, examine its effect beyond the proposing squad.
Start with a decision that crosses teams and describe the minimum all need to uphold. Coordination will be more useful when it allows understanding differences, resolving conflicts, and sharing practices without requiring uniformity where the work needs a different design.
If you want to discuss this decision in the context of your company, talk to dooop.
Further Reading
- How to Prepare a Software Company to Work with AI
- How to Create an Internal Library of Development Examples
- How to Train Developers to Work with AI
Sources
To Continue This Reading
NEXT DECISION
Discuss Application in the Company
Conversation about the software company context
Content by dooop. Registration allows relating this topic to the reader's journey and tracking interest in the subject.