AI agents and automation, explained and practiced

Explain the difference between a chatbot and an AI agent using examples from the work of [your job]. Give one task where a chatbot is enough and one where an agent could help. For the agent example, list the tools it would need and the decisions I should keep for myself.

My goal is [describe the goal]. Break it into numbered steps an AI agent could follow. For each step, say what tool or information it needs, what could go wrong, and whether a human should approve it before the agent continues.

Here is a workflow I want an agent to run: [paste steps]. Mark each step as safe to run automatically, needs my approval first, or should never be automated. Explain each choice in one sentence and focus on actions that cannot be undone.

Write instructions for an AI agent whose job is [describe]. Include: the goal, the tools it may use, the tools it must not use, when to stop and ask me, a limit on how many steps or items it may handle, and how it should report what it did.

Here is my plan for an agent: [paste plan]. Act as a skeptic. List ten ways this could fail, including wrong data, misread instructions, loops, spending too much, and acting on text it found on a web page or in an email. For each, suggest a guardrail.

Here is the log of what my agent did: [paste log]. Check it against my instructions: [paste instructions]. Point out any step that went outside the limits, any result that looks unverified, and anything I should review or undo.

An agent will help me with [task] and may need access to [list apps or accounts]. For each, tell me the smallest permission that would work, whether to use read-only access, and what could go wrong if the access were misused.

I plan to run an agent that [describe the task] about [how often]. Explain what drives the cost of running agents, such as the number of steps, the amount of text read and repeated attempts. Suggest ways to limit it and a simple check I can do after the first week.