Q3Agents & RAG

Q3. Simple workflow or autonomous multi-agent system?

When to use a single LLM call, a deterministic workflow, an agent graph or a multi-agent system, and the guardrails (step limits, token budgets, checkpoints, human handoff) that keep agents from running away.

Rajendra Kumar · · 2 min read

Q3. Simple workflow or autonomous multi-agent system?

I always start with the simplest thing that works. Autonomous agents are exciting, but most business problems have clear steps, and clear steps should be code, not a model deciding what to do next.

A ladder from single LLM call to workflow to agent graph to multi-agent, with workflow as the default. The agent graph connects to four guardrails: max steps, token budget, Postgres checkpoints and human handoff after two failures.
Architecture · Move right only when the problem forces you to. Agent graphs always ship with all four guardrails.

The ladder

  • Single LLM call. Summarise, classify, extract. Most features stop here.
  • Deterministic workflow. Step A, then B, then C, with the LLM doing one job inside each step. My default.
  • Agent graph. The path depends on what the model finds, like triaging a messy customer issue. I use LangGraph with typed state.
  • Multi-agent. Only when work really splits across separate domains or needs parallel research, and the path can't be known upfront.

Agents never get an open-ended loop

  • Step limit. A hard cap on how many steps the graph can take.
  • Token budget. A circuit breaker that stops a run once it costs too much.
  • Checkpoints. State is saved in Postgres, so a run can resume or be inspected.
  • Human handoff. If a tool call fails validation twice, the graph routes to a person instead of retrying forever.

Key takeaways

  1. 01Default to a workflow; earn your way to agents.
  2. 02Every agent run has a step cap and a cost cap.
  3. 03Failure means escalation, not infinite retries.
AI agentsLangGraphMulti-agentWorkflowsInterview prep