A historic cider-mill gift shop wants a helper that checks inventory, reserves a tasting, and emails a confirmation. A single generated paragraph is not enough. Which application pattern fits?
Select an answer to reveal the explanation.
Short Explanation
Think of check inventory, reserve a tasting, email the confirmation. That is an agent chaining tool calls. A one-shot paragraph does not finish the chain.
Full Explanation
Checking inventory, reserving a tasting, and emailing a confirmation is a multi-step workflow that needs tools and orchestration—an agent that can chain those tool calls. Each step depends on prior results (availability before reserve, confirmation details before email), so a single free-text paragraph cannot complete the job reliably. Agent patterns plan, call APIs or functions, observe results, and continue until the gift-shop outcome is done.
A one-shot completion with no tools and no follow-up steps can only emit text; it cannot mutate inventory or send mail. A dashboard that charts last year's tastings is analytics, not transactional booking. Speech-to-text alone converts audio to text; it neither reserves inventory nor chains confirmations.
Exam caveat: agents increase power and also risk—tool permissions, loops, and partial failures matter. Guardrails and clear tool scopes should accompany any booking agent that can touch inventory or customer email.
Operational check: in a sandbox, run a scripted tasting request that forces a stock-check failure then a success path; verify the agent stops or recovers correctly, never double-books, and only sends email after a confirmed reservation ID exists.