dooopSoftware · Organization · 5 min

How to Assess Adoption of AI Development Practices

Observe real behaviors to assess adoption of AI practices. Identify obstacles and revise agreements before increasing enforcement.

Published on September 6, 2026

CENTRAL THESIS

A declaration of adoption does not show whether the practice helps complete the work.

Observation must distinguish lack of understanding, obstacles, and inadequate agreements.

Assessing adoption of new AI development practices means observing whether the agreed behaviors appear in the work and which obstacles prevent their execution. Having access to the tool, attending training, or declaring support for the change alone does not demonstrate consistent use of the practice. Choose a verifiable behavior, examine real tasks, and use the evidence to decide if the agreement needs guidance, better conditions, or revision.

Define the Behavior Before Measuring Adoption

"Using AI responsibly" is a broad intention. Turn it into something observable: recording the task's intention, checking out-of-scope changes, performing a relevant verification, or explaining why a suggestion was rejected.

Also describe when the behavior applies. A practice suitable for code changes may not serve an exploratory task. The person needs to recognize the situation in which the agreement should guide their work.

The DORA on learning culture proposes treating learning as an organizational investment. In adoption assessment, this suggests examining the conditions for practicing, not just requiring a compliance declaration.

Choose Evidence Proportional to the Decision

Use artifacts already part of the workflow when they are sufficient: task description, review record, test evidence, or documented decision. Avoid creating extensive data collection without explaining what decision it will help make.

An observation can verify:

Whether the necessary context was available.
Whether the practice was applied in the intended situation.
Whether the person can explain its purpose.
Whether the produced evidence helps the work recipient.
Whether there was an exception and how it was decided.

Maintain contextual reading. The absence of a record may represent forgetfulness, difficulty of use, lack of time, or format inadequacy. Initial observation should help distinguish these possibilities without anticipating a conclusion about individual attitude.

Investigate Obstacles Before Increasing Enforcement

Discuss a concrete task. Ask when the practice would help, what hindered its application, and which part seems unnecessary. Compare the report with the observed workflow.

If the agreement requires a verification that cannot be performed in the available environment, intervention should address that condition. If the record is long and does not guide any decision, it may need simplification. If the person does not understand the criterion, the next action may be accompanied practice.

The response does not need to be the same for the entire team. A shared process problem calls for a different correction than a localized doubt. Record the cause the intervention intends to address to evaluate the result later.

Fictional Example: A Review Record Does Not Appear

Imagine a team that agreed to record which agent suggestions were rejected before opening the review. When examining tasks, leadership notices the field is empty in some cases.

A conversation reveals that some tasks had no rejected suggestions; in others, the decision was recorded elsewhere; in others, the person did not know what should be explained. Treating all cases as lack of adoption would hide different needs.

The team can adjust the field to distinguish absence of rejection, reference to existing records, and decisions needing justification. Then, observe whether the material became more useful for the review. The example does not presume improvement: it presents a way to formulate and verify the intervention.

Differentiate Formal Adoption and Practice Usefulness

A practice may appear in all forms and still not help the work. Observe whether the information is used by reviewers, whether doubts are resolved at the expected moment, and whether the activity produces the intended effect.

The Microsoft ExP describes the connection between hypothesis, measurement, and iteration. For process change, maintain this link: what difficulty does the agreement intend to reduce and how will the team perceive if the intervention helped?

Do not attribute any change in deadline or quality to the practice without considering other changes during the period. The assessment may indicate a next adjustment without proving a broad causal relationship.

Review the Agreement and Communicate the Decision

When concluding the observation, choose between maintaining the practice, simplifying it, offering support, correcting an execution condition, or removing it. Explain the reason and what the team should do from then on.

The DORA on documentation highlights clarity, location, and reliability. Update the agreement where people consult it. A change discussed in a meeting but absent from the current record leaves the team without a common reference.

Start with one practice and a sample of tasks that help understand it. Adoption assessment should produce a decision on how to improve the work, with sufficient evidence to distinguish lack of understanding, operational obstacle, and rule needing revision.

If you want to discuss this decision in your company’s context, talk to dooop.

Sources

To Continue This Reading

How to Handle Technical Disagreements in AI-Assisted Decisions

NEXT DECISION

Discuss Application in the Company

Discussion about the software company context

Content by dooop. Registration allows linking this topic to the reader’s journey and tracking interest in the subject.