Individual
The person tries AI on small tasks and compares the result with their own judgment, without a stable pattern yet.
Exploring: AI is being tried in limited cases.
You are learning where AI helps, where it misleads, and what should remain human. The experiments are useful, but use is still discretionary and controls are not yet repeatable.
How Exploring appears at each level
Team
People are using AI in pockets and discovering where shared inputs, review, and handoff standards are needed.
Organization
Pilots are underway. The organization is learning where policy, workflow, and risk boundaries need to exist.
Read this as a condition in a particular workflow, not as a permanent label.
Evidence to look for
Problem framing
Experiments have a defined task, expected outcome, and reason AI might be useful.
Grounding and evidence
Users are testing which sources, context, and constraints improve results.
Validation and judgment
First checks are emerging, but the depth of review depends on the person or situation.
Workflow integration
AI assists selected steps, while the surrounding workflow remains largely manual.
Ownership and communication
People recognize that a human remains accountable, but disclosure and escalation are inconsistent.
Learning and improvement
Attempts are compared informally to learn what helps, fails, or should remain human.
Practical next move
- Narrow the work to one or two use cases with a clear owner.
- Record the inputs, prompt or method, review checks, and stop conditions.
- Compare the AI-assisted result with the prior workflow and capture correction load.
Common failure mode
Collecting experiments without decision criteria. A pilot should teach you whether to adopt, adapt, or stop a pattern.
Ready to move to Experimenting when...
A bounded pattern can be repeated, its failure modes are known, review is consistent, and handoffs carry enough context for another person to check the work. Continue to Experimenting, or take the assessment.
Use the AAOS stages to turn a promising experiment into a grounded, validated workflow.