By Sagar Shankaran, Founder of CallSphere
CrewAI's hierarchical process auto-spawns a manager that delegates to role-defined workers. We show what works, what breaks (custom manager prompts!), and how it compares to LangGraph's supervisor for production teams.
Key takeaways
TL;DR — CrewAI's hierarchical Crew gives you a manager-worker shape in ~30 lines of YAML. The default auto-manager is fine for prototypes, broken for production — replace it with a custom Manager Agent and detailed delegation instructions before you ship.
A Crew in CrewAI is a group of role-defined Agents executing under one of three Processes: sequential (assembly line), hierarchical (manager delegates), or consensual (vote). The hierarchical process introduces a Manager Agent that reads the goal, decomposes it, dispatches to workers, and synthesizes results.
flowchart TD
GOAL[Crew goal] --> MGR[Manager agent]
MGR -->|delegate task 1| W1[Researcher]
MGR -->|delegate task 2| W2[Writer]
MGR -->|delegate task 3| W3[Editor]
W1 -->|result| MGR
W2 -->|result| MGR
W3 -->|result| MGR
MGR --> OUT[Final output]
Skip CrewAI hierarchical when: you need fine-grained state control (use LangGraph) or your routing logic is too complex for the auto-manager prompt.
CallSphere uses CrewAI Crews offline for batch content work — generating SEO blog drafts, vertical landing-page copy, and outbound email variants. The live voice path stays in LangGraph (lower latency, finer state control).
A typical Crew: Manager Agent → 3 workers (vertical researcher, copywriter, SEO editor). The manager routes based on the brief and reassembles a draft, which is then handed to the reflection critic (separate pattern) before publishing.
Across 37 agents · 90+ tools · 115+ DB tables · 6 verticals, CrewAI accounts for ~6 of the agents (the content team). Pricing: Starter $149 · Growth $499 · Scale $1,499, 14-day trial, 22% affiliate.
Hear it before you finish reading
Talk to a live CallSphere AI voice agent in your browser — 60 seconds, no signup.
from crewai import Agent, Task, Crew, Process
manager = Agent(
role="Editorial Manager",
goal="Coordinate research, drafting, editing for a 1500-word post.",
backstory="20 years running content teams.",
allow_delegation=True
)
researcher = Agent(role="Researcher", goal="Find 5 facts.", allow_delegation=False)
writer = Agent(role="Writer", goal="Draft from facts.", allow_delegation=False)
editor = Agent(role="Editor", goal="Tighten + SEO.", allow_delegation=False)
crew = Crew(
agents=[researcher, writer, editor],
tasks=[Task(description="Produce post on X", agent=manager)],
process=Process.hierarchical,
manager_agent=manager, # custom, NOT auto
verbose=True
)
crew.kickoff()
Process.hierarchical without manager_agent= spawns a generic manager with no domain knowledge. It mis-delegates. Always pass a custom Manager Agent.allow_delegation=True on workers turns the crew into a swarm. Set it False except on the manager.Q: Crew or LangGraph supervisor? Crew for declarative role-based content work. LangGraph for fine-grained state, low-latency, branching logic.
Q: Does Crew handle streaming? Limited. Streaming is per-agent-task; mid-crew streaming is awkward. Use Flow or LangGraph if streaming matters.
Q: How many workers per crew? 3–6 is the sweet spot. Past 8, manager delegation accuracy degrades.
Q: Can I nest crews? Yes — wrap a Crew as a Tool and have a parent Crew call it. Or use CrewAI Flow to orchestrate multiple Crews.
Q: Memory across runs?
Crew has built-in short-term, long-term, and entity memory. Configure memory=True and provide an embeddings model.
There is a clean theory behind crewAI Manager + Workers and there is a messier reality. The theory says agents reason, plan, and act. The reality is that agents stall on ambiguous tool outputs and double-spend tokens unless you put hard limits in place. Once you frame crewai manager + workers that way, the design choices get easier: short tool descriptions, narrow argument types, and a hard cap on tool calls per turn beat any amount of prompt engineering.
Agentic AI in a real call center is a different beast than a single-LLM chatbot. Instead of one model answering one prompt, you orchestrate a small team: a router that decides intent, specialists that own a vertical (booking, intake, billing, escalation), and tools that read and write to the same Postgres your CRM trusts. Hand-offs are where most production bugs hide — when Agent A passes context to Agent B, anything that isn't explicit in the message gets lost, and the user feels it as the agent "forgetting." That's why the systems that hold up under load are the ones with typed tool schemas, deterministic state stored outside the conversation, and a hard ceiling on tool calls per session. The cost story is just as important: a multi-agent loop can quietly burn 10x the tokens of a single-LLM design if you let it think out loud at every step. The fix isn't a smarter model, it's smaller agents, shorter prompts, cached system messages, and evals that fail the build when p95 latency or per-session cost regresses. CallSphere runs this pattern across 6 verticals in production, and the rule has held every time: the agent you can debug in five minutes will out-survive the agent that's "smarter" on a benchmark.
Still reading? Stop comparing — try CallSphere live.
CallSphere ships complete AI voice agents per industry — 14 tools for healthcare, 10 agents for real estate, 4 specialists for salons. See how it actually handles a call before you book a demo.
Q: Why does crewAI Manager + Workers need typed tool schemas more than clever prompts?
A: Scaling comes from constraint, not capability. The deployments that hold up keep each agent narrow, cap tool calls per turn, cache the system prompt, and pin a smaller model for routing while reserving the larger model for synthesis. CallSphere's stack — 37 agents · 90+ tools · 115+ DB tables · 6 verticals live — is sized that way on purpose.
Q: How do you keep crewAI Manager + Workers fast on real phone and chat traffic?
A: Hard ceilings beat heuristics. A maximum step count, an idempotency key on every tool call, and a fallback to a deterministic script when confidence drops below a threshold are what keep the loop bounded. Evals that simulate noisy inputs catch the rest before they reach a real caller.
Q: Where has CallSphere shipped crewAI Manager + Workers for paying customers?
A: It's already in production. Today CallSphere runs this pattern in Sales and IT Helpdesk, alongside the other live verticals (Healthcare, Real Estate, Salon, Sales, After-Hours Escalation, IT Helpdesk). The same orchestrator code path serves voice and chat — the difference is the tool set the router exposes.
Want to see after-hours escalation agents handle real traffic? Spin up a walkthrough at https://escalation.callsphere.tech or grab 20 minutes on the calendar: https://calendly.com/sagar-callsphere/new-meeting.
Written by
Sagar Shankaran· Founder, CallSphere
Sagar Shankaran is the founder of CallSphere, where he builds production AI voice and chat agents deployed across healthcare, hospitality, real estate, and home services. He writes about agentic AI, LLM engineering, and shipping voice agents that handle real calls in production.
See how AI voice agents work for your industry. Live demo available -- no signup required.
How to design a multi-agent system using MCP for tools and A2A for cross-vendor coordination, with a CallSphere voice agent as a participating node.
A2A is the open standard for agent-to-agent coordination. Here is how the Agent Card JSON works, how discovery happens, and what to publish.
A2A unlocks cross-vendor agent coordination, but most enterprise voice/chat workloads still ship faster on a single-vendor stack. Here is how to choose.
Fully autonomous agents are still a fantasy in production. LangGraph's interrupt() lets you pause for human approval mid-graph without losing state. We cover approve/edit/reject/respond actions and CallSphere's escalation ladder.
Enterprise CIO Guide perspective on AutoGen 0.5 brings async-first execution, an extension architecture, and tighter Azure integration.
A Miami DTC brand uses CrewAI Studio to run a daily merchandising agent crew that updates collections, hero copy, and email sends — and the lift it produced.
© 2026 CallSphere LLC. All rights reserved.