Product design question
Design the core interaction model for a Project where a human and multiple agents collaborate on ongoing work. How would you handle delegation, approvals, edit history, permissions, and recovery from bad agent actions so the experience feels powerful but dependable?
- Perplexity
- Product design
- Hard
Practice this question out loud. An AI interviewer asks it, follows up like a real interviewer would, and scores your answer. Type or speak.
Start a mock interview on this question · Mock interview from a job description
What this question tests
Tests designing a multi-agent collaboration interaction model that stays powerful but dependable, covering delegation, approvals, history, permissions, and recovery from mistakes.
How to approach it
- Define the core unit of work: a task delegated to an agent within the Project, with a clear owner, a status, and a required approval step before any external or irreversible action.
- Design delegation: the human assigns a task to an agent with explicit scope, and the agent can request clarification rather than guessing when scope is ambiguous.
- Design approvals: default to requiring human approval for any action outside the Project's workspace, for example sending a message or publishing, with an option to pre-approve low-risk repeated actions.
- Design edit history: every agent action is logged with a diff or summary, tied to the task that triggered it, so a human can review what changed and why.
- Design permissions: scope each agent's access to only the Project's resources by default, with explicit, auditable grants for broader access.
- Design recovery: every agent action must be reversible or have an undo path, and a bad action should roll back to the last known-good state without losing unrelated work.
What a strong answer includes
- Makes approval the default for anything leaving the Project workspace, trading some speed for dependability, with a clear path to pre-approve trusted low-risk actions later.
- Ties edit history directly to the originating task, so review is contextual, not just a flat activity log.
- Scopes agent permissions to the Project by default, treating broader access as an explicit, auditable exception.
- Designs recovery as a first-class requirement, reversible actions and rollback, not an afterthought bolted on after failures happen.
Common mistakes
- Designing for power first and adding approvals or recovery as an afterthought once trust issues surface.
- Building one flat permission level for all agents instead of scoping access per Project.
- Logging actions without tying them back to the task or decision that caused them, making review hard.
Likely follow-up questions
- How would you handle an agent that needs to act outside its Project scope legitimately?
- What happens if two agents make conflicting edits at the same time?
More product design questions
- How would you improve Perplexity's Comet browser for daily use?Perplexity · Product design · Medium
- Design an AI shopping assistant inside Comet.Perplexity · Product design · Medium
- Design a 'Deep Research' feature that produces trustworthy, citable reports.Perplexity · Product design · Hard
- How would you design Comet Assistant to safely act on tasks like booking flights and managing email?Perplexity · Product design · Hard
- Users of Perplexity Computer say they do not understand why the agent clicked, searched, or changed course, and they hesitate to let it continue. Design a core interaction model that makes the agent’s intent, current state, and next actions understandable and controllable without turning the product into a complex dashboard.Perplexity · Product design · Hard
- Perplexity notices that many users start a research task on web, ask several follow-ups, but do not return to that work on another platform later. How would you identify the single highest-leverage friction point in the begin → continue → return journey, and decide whether to invest first in context carryover, navigation, or collaboration?Perplexity · Product design · Hard
More questions from Perplexity
Learn the skill behind it
Chapters of the AI PM course that teach what this question tests.
- Chapter 4: Discovery and strategy for AI products
- Chapter 7: AI UX and human oversight: design for a system that is wrong sometimes
- Chapter 14: Get the job: the AI PM interview loop