AI agent fleet assessment
Which AI agent controls should you check first?
You may know what your agents do. Can your team also say who owns them, what they can access, and who can pause them? Use seven questions to make that conversation concrete. The AI Agent Fleet Risk Calculator turns your answers into a self-assessment score and three prompts for deciding what to verify first.
Seven questions for your next fleet review
Bring these questions to the people who own the agents and the controls around them. For an answer you cannot confirm, identify the record or workflow that would settle it.
- Fleet size: Which agents are in use, and is your inventory current? Check the list against the agents your team runs in production and who maintains that list.
- Governance owner: Who owns each production agent and changes to its role? Find the named owner and the path for approving a change.
- Access and permissions: Who can grant, change, or revoke an agent's access? Review the permissions it has and how access is removed.
- Production monitoring: What would show unexpected behavior, and who sees it? Identify the signals people actually review and where they go.
- Incident response: Who responds if an agent needs attention, and how can it be paused? Locate the response procedure and the person authorized to act.
- Decision records: Can you trace a consequential agent decision? Find the record a reviewer would use and who is responsible for keeping it.
- Human approval: Which consequential actions need a person's approval? Compare the rule with the approval step in the real workflow.
You can use this list on its own. The assessment puts your selected answers into a starting order for further review.
What the assessment gives you
Choose one answer for each of the seven areas. The tool combines them into a weighted 0–100 self-assessment score. Under this rubric, higher scores reflect answers assigned more points across fleet size and six governance areas. Before entering an email, you see the score and three signals drawn from your answers. Treat each signal as a question to check with your team.
Submit an email in the tool to open the detailed report on screen. It shows a score for each of the seven areas and four actions ranked from your answers. That ranking offers a starting order for discussion; your team decides what needs attention after checking the actual controls.
The score reflects your selected answers. It does not inspect your systems, verify how well controls work, or predict an incident.
Turn one uncertain answer into a next step
Pick the question you were least sure how to answer. Name the person who can confirm it, then look at the relevant permission, record, procedure, or approval step together. You now have a specific control and owner for the next conversation.
Frequently asked questions
What does a higher score mean?
Your answers received more points under this weighted rubric, which includes fleet size as well as six governance areas. The number is a starting point for review, not a comparison with other fleets.
Does the assessment scan my agents or systems?
No. You select answers about your fleet and its controls; your team checks the underlying systems and records.
What can I see before entering an email?
After all seven answers, you see your self-assessment score and three signals based on those answers.
What happens after I enter an email?
After you submit an email successfully, the tool opens a detailed report on screen with seven area scores and four ranked actions.
Start your fleet review
Use the assessment to put seven answers into a starting order, then check the controls behind them with your team.