Model answers for 4 agentic AI questions
Typical round: concepts, a design exercise, a project deep dive
Q: What makes a system 'agentic'? A: The model decides its next step instead of following a fixed script: it plans, calls tools, reads the results and continues until a goal or stop condition is met. A plain chatbot or a single prompt chain is not an agent.
Q: How does tool calling work? A: You describe tools with names, descriptions and JSON schemas. The model returns a structured call, your code executes it, and the result is passed back as a message. Clear tool descriptions and strict input validation matter more than clever prompts.
Q: How do you stop an agent from looping or overspending? A: Set step and token budgets, detect repeated actions, add timeouts per tool, and require human approval for irreversible or expensive actions. Log every step so failures can be replayed.
Q: When would you use multiple agents instead of one? A: When tasks split cleanly into specialised roles with separate context, like a researcher and a writer, or when work can run in parallel. Otherwise a single agent with good tools is simpler, cheaper and easier to debug.