Skip to content
Last updated

Using the Marketing Super Agent

The Marketing Super Agent acts as your central orchestrator. Instead of manually selecting individual AI agents for specific tasks, you provide a high-level goal, and the Super Agent autonomously selects the optimal combination of "Task Agents" (specialists) to execute the work.

This allows you to complete complex, multi-step marketing and sales campaigns within a single, unified workflow(chat)

Orchestrating a Workflow

To use the Marketing Super Agent, you will define a goal, review the proposed execution plan, and generate results.

1. Define Your Goal

  1. In the chat interface, input your specific goal or task. Be as detailed as possible regarding the target audience, context, and desired outcome.

    Example Prompt: "For the new product A, identify the potential psychological resistance of the target persona (woman in her 30s, time-saving oriented) and devise a campaign concept (based on behavioral economics) to overcome that resistance."

2. Review the Proposed Plan

Once you submit your goal, the Orchestrator analyzes the request and leverages internal Plan Templates to propose a workflow.

  1. Review the Plan Proposal displayed in the chat.
  2. Verify the specific agents selected and the order of execution.

plan sample

  1. If the plan looks correct, type "Continue" or a similar confirmation.
  2. If adjustments are needed (e.g., you want to add a specific constraint), describe the change in natural language.

3. Execution and Output

The Orchestrator sequentially calls the necessary Agents. The output of one agent (e.g., research data) is automatically passed as input to the next agent (e.g., strategy generator).

  • Wait for completion: Complex chains may take time as web searches (powered by OpenAI's GPT-5 mini) are performed.
  • Download Results: Final text outputs are displayed in the chat window.

Specifying Output Formats

Some agents are capable of generating specific file formats, such as PowerPoint slides (.pptx).

  • Requirement: You must specify the desired output format before the task begins (in your initial prompt).
  • limitation: Changing the output format after results are generated usually requires re-executing the entire sequence.

Common Use Cases

The following table outlines common goals and how the Super Agent maps them to specialized sub-agents.

GoalPrimary PlanKey Specialist AgentExpected Outcome
Competitor Analysiscounter_strike_strategycompetitor_campaign_researcherAnalysis of competitor campaign intent and differentiation strategies.
Sales Proposal Reviewdeal_closer_proposalb2b_proposal_sounding_boardPre-review of sales documents acting as a customer decision-maker with improvement advice.
Value Translationfeature_killer_pitchfeature_benefit_translatorTranslation of technical specs into emotional "Killer Values" for marketing copy.

Troubleshooting & FAQ

Q. The intended agent is not being called.

If the Orchestrator does not select the agent you expect, try specifying the agent explicitly in your prompt (e.g., "Use the B2B Proposal Sounding Board to..."). You can also ask the orchestrator to "provide a list of agents" to see what is currently registered.

Q. Can the agent create charts or segments?

No. The Marketing Super Agent does not support chart creation, segment creation, or Plotly image generation.

Q. Which AI model is used?

  • The orchestrator and task agents use Cloud Haiku 4.5 or Sonnet 4.5.
  • Web search capabilities are powered by the OpenAI GPT-5 mini API.

Q. Can the agent access and analyse my dataset in treasure data CDP?

Not yet. The Marketing Super Agent does not support to access CDP data yet.