Would you allow an AI agent to deploy your app? What would it need to prove?

I’m interested in knowing the boundary between an AI agent that can prepare a change and one that can make the change.

For a personal experiment I may be ok with a lot of automation. However if it’s a tool for my team and touches customer data/production systems I would want a clearer story around permissions, review and recovery.

Where would you draw the line?

• Reading logs/repo – Ok without asking?
• Opening a pull request – Ok after tests?
• Deploy to staging – Automatic or confirmed?
• Deploy to production – What evidence would you want?
• Changing data/external action – Always confirmed? Are there scenarios where you would automate?

I’m not looking for one size fits all. There is a different risk for a demo, internal tool and production.

If you have created an agent workflow, what boundary have you set and why?

discuss #ai #devtools