AgentforceArchitect (Hardest)multi-agenthelpdeskdesign
Compare single monolithic agent vs multi-agent for employee IT helpdesk
Real World Scenario
IT wants one Agentforce employee agent for password reset, laptop requests, and HR policy questions; platform team proposes specialized workers.
Expected Answer
• Monolith simpler initially but prompt and action sprawl degrades accuracy
• Specialized workers with narrow knowledge reduce hallucination and permission scope
• Supervisor routes by intent classification with shared session context
• Permission boundaries easier per domain agent (HR vs IT systems)
• Start monolith with modular topics; split when containment drops below SLA
• Cost: multiple agent invocations vs single long context
• User experience: seamless handoff messaging between workers
Follow-Up Questions & Answers
Click to expand — each follow-up includes a direct, interview-ready answer
Direct answer: Monolith simpler initially but prompt and action sprawl degrades accuracy Also consider: Specialized workers with narrow knowledge reduce hallucination and permission scope In practice: Supervisor routes by intent classification with shared session context Document the decision in an ADR and align with enterprise standards.
Architect Perspective
Split agents on permission and knowledge boundaries not org chart lines alone.