The 5-Claude Media Buying Team

Most media buyers do not need one smarter AI chat. They need a better division of labor. One general Claude can help with a lot of paid media work:

• campaign diagnosis
• search terms
• creative fatigue
• weekly reporting
• budget pacing
• competitor teardown
• lead quality

But after a few turns, one general chat usually starts blending jobs together. The analyst starts writing creative. The creative strategist starts recommending budget moves. The media buyer starts guessing tracking issues. That is how you get confident advice from a messy context window. The cleaner setup is to split the work into five specialist Claude roles. Each role has:

• a narrow job
• specific inputs
• a fixed output format
• a missing-data check
• a human approval gate

Here is the team.

1. Media Buyer

The Media Buyer owns account structure, pacing, budget movement, and decision priority.

Use this role when you need to answer:

• where is spend going?
• what should move this week?
• what is pacing off-target?
• what needs approval before changing spend?
• what is the single most important account issue?

Starter prompt.

You are my Media Buyer.

Your job is to turn paid media account data into a short decision memo.

Use this context:
- business goal:
- target CPA / ROAS:
- main campaigns:
- current budget:
- last 30 days:
- last 7 days:
- known constraints:

Separate facts from hypotheses.
Do not recommend budget changes unless you explain the risk, confidence, and approval needed.

Output:
1. what changed
2. why it matters
3. where spend is leaking
4. what to hold
5. what to fix
6. what needs approval

2. Google Ads Specialist

The Google Ads Specialist owns search terms, keyword quality, auction pressure, PMax questions, and Google-specific diagnosis.

Use this role when you need to answer:

• where is search spend leaking?
• are the clicks relevant?
• is Quality Score hiding a repair queue?
• is performance limited by budget, rank, or signal quality?
• is Google Ads optimizing toward the right conversion?

Starter prompt.

You are my Google Ads Specialist.

Review the pasted Google Ads data and find the highest-confidence issues first.

Inputs:
- campaigns:
- keywords:
- search terms:
- cost:
- conversions:
- conversion value:
- search impression share:
- lost IS budget:
- lost IS rank:
- Quality Score / expected CTR / landing page experience:

Rules:
- Do not blame bid strategy before checking conversion signal quality.
- Separate search intent problems from landing page problems.
- Mark every recommendation as do_now / investigate / monitor.

Output a repair queue with evidence and missing data.

3. Meta Ads Specialist

The Meta Ads Specialist owns creative fatigue, frequency, audience overlap, learning status, and delivery patterns.

Use this role when you need to answer:

• is this creative fatigue or weak offer?
• is frequency creating pressure?
• are ad sets fighting for the same audience?
• did performance change because of delivery, tracking, or landing page issues?
• are cheap leads actually good leads?

Starter prompt.

You are my Meta Ads Specialist.

Review this Meta Ads export and explain what is likely driving performance.

Inputs:
- campaign objective:
- spend:
- impressions:
- reach:
- frequency:
- CPM:
- CTR:
- CPC:
- conversions / leads:
- CPA:
- creative names:
- last 7 vs last 30 days:

Rules:
- Do not say “test more creatives” unless you show what creative signal supports it.
- Separate fatigue, audience, offer, tracking, and landing page hypotheses.
- Flag missing evidence clearly.

Output:
1. top 3 likely issues
2. evidence for each
3. evidence against each
4. next check
5. recommended action

4. Creative Strategist

The Creative Strategist owns winning angles, competitor patterns, ad-to-page match, and creative briefs.

Use this role when you need to answer:

• what angle should we test next?
• what promise is the ad making?
• does the landing page keep that promise?
• what objections are showing up in comments or sales calls?
• what competitor pattern is worth stealing or avoiding?

Starter prompt.

You are my Creative Strategist.

Turn performance data, comments, objections, and competitor ads into creative briefs.

Inputs:
- best performing ads:
- worst performing ads:
- landing page:
- offer:
- customer objections:
- competitor examples:
- performance notes:

Rules:
- Do not give generic “hook ideas.”
- Every angle must name the pain, promise, proof needed, and risk.
- Separate performance-backed angles from hypotheses.

Output 10 angles, then pick the top 3 to brief.

5. Performance Analyst

The Performance Analyst owns weekly reporting, metric changes, anomaly explanation, and stakeholder clarity.

Use this role when you need to answer:

• what changed this week?
• what is fact vs hypothesis?
• what should the founder, client, or operator know?
• what needs approval?
• what are the three decisions?

Starter prompt.

You are my Performance Analyst.

Turn this weekly paid media data into a decision-ready update.

Inputs:
- previous period:
- current period:
- spend:
- conversions:
- CPA / ROAS:
- conversion value:
- channel split:
- campaign notes:
- known tracking issues:

Rules:
- Do not bury the recommendation.
- Separate facts from hypotheses.
- End with decisions, not dashboards.

Output:
1. what changed
2. what likely caused it
3. what is still unclear
4. three priorities
5. owner approvals needed

How to set it up

You do not need a complex multi-agent system to start.

Start with five saved prompt files or snippets:

1. media-buyer.md
2. google-ads-specialist.md
3. meta-ads-specialist.md
4. creative-strategist.md
5. performance-analyst.md

Then give each role.

• the account context it needs
• the relevant export or screenshot
• the business goal
• the target CPA / ROAS if available
• the decision you want
• the approval boundary

Do not ask every role to solve everything. That is the whole point. The Google Ads Specialist should not rewrite your weekly report. The Creative Strategist should not move budget. The Performance Analyst should not invent creative angles without evidence. The Media Buyer keeps the decision layer.

Why this works better than one Claude

One general Claude tries to be helpful everywhere. That is useful for brainstorming. It is weaker for paid media diagnosis. Specialized roles work better because:

• the context stays narrower
• the output format is predictable
• missing evidence is easier to catch
• you can run checks in parallel
• the final recommendation has a clear owner
• spend and tracking changes still need human approval

The goal is not to let AI run the ad account.

The goal is to stop using AI like a messy assistant and start using it like a small operator team.

Guardrail

Do not let these roles:

• change budgets automatically
• pause campaigns automatically
• edit conversion tracking automatically
• publish ads automatically
• message customers automatically
• pretend hypotheses are facts

Use them to prepare better decisions.

Then approve the decisions yourself.