One client campaign.
A team you can follow.

A small agency wants to turn an approved service offer into a landing page and three social posts. The owner has two hours to review the first week’s work.

This is a fictional example, not a customer case study. It shows how to divide a campaign workflow; your own brief may justify fewer agents.

The focus: one campaign, not the whole agency

Keep research, content and measurement within the same campaign. This example does not require three business areas just because three specialists take part. A separate sales or operations workflow would need its own scope and handoffs.

Who does what?

You: approve the offer and angle, check the drafts and publish. The AI Team Lead coordinates the work; it does not take your approval role.

Set it up in this order

1. Create your real company brief in Agent Team Plan and confirm the campaign as its main focus.
2. Generate your blueprint. In Grok Bot or your chosen AI platform, create the recommended agents and paste each agent’s own setup instructions.
3. Give the Team Lead the approved brief and the starting message below. Adapt its deadline and deliverables first.
4. If your platform does not pass work between agents, copy the Team Lead’s task into the named specialist’s chat, then bring the answer back to the Team Lead.

When and where to give the goal

Send this to the Team Lead after setup, as its first task message. The goal travels with the assignment; you do not need to invent a separate target for each agent.

This message assumes the four roles shown above. If your generated blueprint recommends a different team, use its role names and starting message instead.

You are the team lead for this fictional agency pilot. Before starting, ask me to confirm the real client, approved offer, target audience, source documents, review owner and start date. Do not treat this example as facts about my business.
Goal: within seven days of the agreed start date, produce one source-backed campaign brief, one landing-page draft and three supporting social drafts for human review.
First ask the Research agent for evidence and unanswered questions. Bring its proposed angle to me for approval. Only then give the approved brief to the Content agent. Ask the Measurement agent to prepare a simple results sheet using available data; mark missing measurements as unknown.
Return one progress update with the next decision I need to make. Do not publish, send messages, change client accounts or spend money. If you cannot communicate with another agent, give me the exact message to paste and tell me where to bring its answer. Stop after one revision round and ask me what to change next.

Tools and access: start with approved files

• Research: approved client documents and website excerpts. A read-only document connection is optional; pasting the relevant text is enough to begin.
• Content: a draft document or your AI chat. Website publishing access is unnecessary for this pilot.
• Measurement: a spreadsheet of dates, review time and available results. If you already use Google Analytics, its Viewer role can read reports. In Search Console, a Restricted user can view most data. You can also export and share selected reports yourself.

Connect tools in the AI platform where your agents run. Agent Team Plan does not grant these permissions. Check the connector’s actual scope before approving it. Sources: Google Analytics access roles ↗ and Search Console permissions ↗.

What counts as a useful first week?

By day seven, you can review one brief, one landing-page draft and three social drafts, with sources and open questions visible. Record how much correction and review time they needed. If you later publish them, agree a separate measurement period; seven days of drafting cannot establish a sales improvement.

Update the brief when the audience, offer or campaign priority changes. Generate a new version only when you want revised agent instructions, then apply them in your AI platform.

Compare packages → · See the one-agent consultant example →