CrewAI marketing crew: reviewer defined, not wired

Why it matters

This is a public, end-to-end marketing team spec: a Lead Market Analyst researches, a Chief Marketing Strategist plans, and a Creative Content Creator writes, each step returning typed JSON. A fourth role sits in the config: the Chief Creative Director, whose goal is to review, approve, ask clarifying questions or delegate follow-up work. In the default run that reviewer never appears, because crew.py sets a sequential process and leaves the hierarchical line commented out. The fix is documented, not hidden: run the hierarchical process with the director as manager agent. One caveat on age: tasks.yaml still says the current year is 2024, so treat this as an older official example.

Roles and reporting lines

Role	Position in the run	Tools	Tasks
Inputs (human)	Start	-	`customer_domain`, `project_description`
Lead Market Analyst	Step 1	web search, scrape website	research_task
Chief Marketing Strategist	Step 2	web search, scrape website	project_understanding_task, marketing_strategy_task
Creative Content Creator	Step 3	-	campaign_idea_task, copy_creation_task
Chief Creative Director	Defined in agents.yaml, not instantiated in crew.py	-	Goal: oversee the team's work; review, approve, ask clarifying questions or delegate follow-up

Routines

One run per brief, in order: research → project understanding → marketing strategy (JSON: name, tactics, channels, KPIs) → 5 campaign ideas (JSON) → copy (JSON: title, body)

Every step returns typed JSON, so the output of one role arrives in a fixed shape that the next role can check.

Approval gates

Where	Status	Source
Default run	No review step; `process=Process.sequential`, `# process=Process.hierarchical` commented out	crew.py
To add the gate	Hierarchical process, Chief Creative Director as `manager_agent`	CrewAI docs, Hierarchical Process
Human sign-off	Not part of the example	-

Copyable spec

org_chart:
  name: Marketing strategy crew
  number: 3
  platform: CrewAI (official example repo)
  source: https://github.com/crewAIInc/crewAI-examples/tree/main/crews/marketing_strategy
  credit: crewAIInc, README 'By @joaomdmoura'; MIT license (per README)
inputs: [customer_domain, project_description]
roles:
  - {id: analyst, title: Lead Market Analyst, tools: [web search, scrape website], tasks: [research_task]}
  - {id: strategist, title: Chief Marketing Strategist, tools: [web search, scrape website], tasks: [project_understanding_task, marketing_strategy_task]}
  - {id: creator, title: Creative Content Creator, tasks: [campaign_idea_task, copy_creation_task]}
  - {id: director, title: Chief Creative Director, status: "defined in agents.yaml, not instantiated in crew.py", job: "oversee the team's work; review, approve, ask clarifying questions or delegate follow-up"}
task_order: [research, project understanding, strategy (JSON: name, tactics, channels, KPIs), 5 campaign ideas (JSON), copy (JSON: title, body)]
process: sequential   # crew.py comments out process=hierarchical
to_add_the_gate: "use the hierarchical process with the director as manager_agent (CrewAI docs: manager delegates and validates outcomes)"
approval_gates:
  - {where: final copy, approver: none in default run}
  - {where: human, documented: false}

About the maker

CrewAI Inc. is the company behind the CrewAI framework, an open-source tool for building teams of AI agents: each agent gets a role, a goal and a backstory, agents are grouped into a crew, and the crew runs its tasks in sequence or under a manager.

Next to the framework, the company sells an enterprise platform for putting agents into production, in four parts as presented on crewai.com:

Discovery, which scans a company's tickets, chats and workflows to rank what to automate first;
Build, a visual studio and a code API for creating the agents;
Govern, a control plane with tracing, access control and human approval gates in the execution path;
Optimize, which evaluates production runs and uses them to improve the agents over time.

This marketing crew is one of the worked examples the CrewAI team publishes to show how a crew is put together. It demonstrates the framework; it is not evidence of how the company runs its own marketing.

crewai.com/open-source · LinkedIn company page

Sources

crewAIInc, crewAI-examples: marketing_strategy, README by @joaomdmoura. Folder fetched; agents.yaml, tasks.yaml, crew.py read raw on 15 Oct 2026, 22:14 to 22:16 CEST.
CrewAI docs, Hierarchical Process. Fetched 15 Oct 2026, 22:14 CEST.
Maker: CrewAI Inc., crewai.com/open-source, the framework's home; company page on LinkedIn. Checked 18 Oct 2026.

What is confirmed / what is not

Confirmed:

agents.yaml: 4 roles, incl. chief_creative_director with the review-and-approve goal
crew.py: agents for 3 roles only; sequential process; hierarchical line commented out
tasks return typed JSON (MarketStrategy, CampaignIdea, Copy)
CrewAI docs: manager agent and manager_agent

Not confirmed:

tasks.yaml says "the current year is 2024": older official example, not necessarily current best practice
README states MIT; GitHub detects no license file in the repo
README's default model left out on purpose

All org charts · All bot templates

The reviewer is in the config, but you have to wire it in.