Pattern: Parallel Workers | Team size: 6 agents
This team attacks performance improvements by running specialized optimization streams in parallel across Google and Meta. Parallelization speeds iteration on creatives, audiences, bidding, and landing pages while a coordinator consolidates learnings into a single roadmap.
Automatization of google & meta ads management
Create an agent team to automate Google & Meta ads management end-to-end, improving efficiency (CPA/ROAS), iteration speed, and governance by running parallel optimization streams across account structure, delivery, creative testing, audiences/CRM, and landing page CRO, then consolidating into a single execution roadmap. PROJECT NAME (use exactly): ads_management_automation GLOBAL CONTEXT (read carefully; all agents must follow) - Objective: design an automation-ready operating system for Google Ads + Meta Ads management that can be implemented by an engineering/ops team, including rules, data flows, naming conventions, testing frameworks, dashboards, and a prioritized rollout plan. - Scope includes: campaign/account structure, bidding/delivery guardrails, search query/negative automation, creative testing pipeline, audience segmentation + CRM/offline conversions, landing page experiment plan, measurement + attribution, alerting, and change-log governance. - Deliverables must be written so a performance marketing team and an implementation team can act on them without further clarification. - Assume the company has: Google Ads + Meta Ads accounts, GA4, Google Tag Manager, a CRM (e.g., HubSpot/Salesforce), and a server-side option (CAPI / enhanced conversions) available. Do not assume any other tools; if you recommend tools, present them as optional. - Use consistent terminology and enforce it across documents: “automation rule”, “guardrail”, “test”, “promotion”, “rollback”, “change log entry”. - All outputs must be saved under: outputs/agent_teams_demo/ads_management_automation/ - Writing constraints: use clear headings, numbered steps, and tables where helpful. Be explicit about what is automated vs manual, and what data is required. - Collaboration rule: each specialist must produce (1) recommendations, (2) automations/rules with thresholds, (3) required inputs/data, (4) risks/failure modes, (5) how to measure success, and (6) dependencies on other streams. - Coordination rule: if an agent’s recommendation conflicts with another stream (e.g., budget allocation, overlapping audiences, creative fatigue vs learning phase), explicitly flag the conflict in their file and propose a resolution option. TEAM DESIGN: Growth Lab Swarm (Parallel Workers) Agents (run in parallel after coordinator sets shared assumptions): 1) Google Ads Optimization Specialist 2) Meta Ads Optimization Specialist 3) Creative Testing Strategist 4) Audience & CRM Integrations Specialist 5) Landing Page & CRO Analyst 6) Swarm Coordinator PHASE 0 — COORDINATOR SETUP (MUST COMPLETE BEFORE ANY OTHER AGENT BEGINS) Swarm Coordinator: - Create a shared “Assumptions & Definitions” doc that all agents will use. - Define a baseline account taxonomy and naming convention scaffolding that everyone must reference. - Define success metrics and guardrails that apply across channels. Coordinator deliverable (must be completed first): 1) outputs/agent_teams_demo/ads_management_automation/00_assumptions_definitions.md Requirements (800–1200 words, use these sections in this order): - Business Goal & Primary KPIs (CPA/ROAS/MER, volume, LTV if applicable) - Conversion Taxonomy (Primary vs Secondary conversions; micro/macro; MQL/SQL/Sale examples) - Measurement Stack (GA4, GTM, Ads pixels, CAPI/enhanced conversions; what “source of truth” is) - Attribution & Reporting Principles (what decisions use which attribution view) - Global Guardrails (spend volatility limits, learning-phase protections, minimum data thresholds) - Naming Conventions v0 (campaign/ad set/ad/asset naming components; required tags) - Change Governance (who approves, rollback expectations, cadence) PHASE 1 — PARALLEL SPECIALIST WORKSTREAMS (BEGIN ONLY AFTER 00_assumptions_definitions.md EXISTS) Each specialist produces exactly one primary markdown file at the path specified below. Each file must be 1200–1800 words and must include the required sections exactly as listed. 1) Google Ads Optimization Specialist File: outputs/agent_teams_demo/ads_management_automation/10_google_ads_automation_plan.md Sections (in this order; include tables where indicated): - Current-State Risks to Efficiency (structure, match types, queries, tracking, PMax/Shopping caveats) - Target Account Structure Blueprint (campaign types; when to separate brand/non-brand; geo/device; table: Campaign Type → Purpose → KPI → Notes) - Automation Rules & Scripts (table required) Table columns: Rule Name | Applies To | Trigger/Threshold | Action | Frequency | Safeguards | Rollback Include at least: search query mining + negatives, budget pacing, bid strategy guardrails, RSA asset hygiene, geo/location exclusions, dayparting (if applicable), PMax asset group management guardrails, anomaly detection alerts. - Keyword/Query Management System (exact workflow: what is automated vs manual; query-to-keyword promotion criteria) - Bidding & Budget Policy (how to choose tCPA/tROAS/max conversions; minimum conversion thresholds; learning phase protection) - Measurement Dependencies (enhanced conversions, offline conversions, GA4 import caveats) - Failure Modes & Mitigations (at least 8 bullet points) - Success Metrics & Weekly Review Checklist (10–15 checklist items) 2) Meta Ads Optimization Specialist File: outputs/agent_teams_demo/ads_management_automation/11_meta_ads_automation_plan.md Sections (in this order): - Current-State Risks to Performance (audience overlap, learning limited, placement issues, tracking gaps) - Campaign Architecture Blueprint (Advantage+ vs manual; prospecting vs retargeting; table: Campaign → Objective → Audience → Creative Type → KPI) - Automation Rules (table required) Columns: Rule Name | Level (Campaign/Ad Set/Ad) | Trigger/Threshold | Action | Frequency | Safeguards | Rollback Include at least: budget scaling rules, creative fatigue detection, placement guardrails, audience expansion toggles, cost cap/bid cap guidance, pausing rules with minimum spend thresholds, anomaly alerts. - Delivery Optimization Policy (how to handle learning phase, consolidation vs segmentation, frequency management) - Measurement Dependencies (Pixel + CAPI, aggregated events, event prioritization, offline events) - Failure Modes & Mitigations (at least 8 bullet points) - Success Metrics & Weekly Review Checklist (10–15 checklist items) 3) Creative Testing Strategist File: outputs/agent_teams_demo/ads_management_automation/12_creative_testing_system.md Sections (in this order): - Creative North Star (message-market fit hypotheses; offer angles; what “winning” means by funnel stage) - Test Design Framework (how to structure tests for Google RSAs/PMax assets and Meta ads; isolate variables; learning requirements) - Creative Test Matrix (table required) Columns: Funnel Stage | Hypothesis | Variable | Variants (n) | Channels | Primary Metric | Min Sample | Decision Rule | Promote Path - Naming Conventions (must extend Coordinator v0; specify mandatory tags: hook/angle/format/creator/offer/date) - Creative Lifecycle Automation (intake → QA → launch → monitor → promote → retire; what can be automated; fatigue rules) - Asset QA Checklist (20-item checklist; include compliance/brand, landing page parity, tracking parameters) - Reporting & Winner Declaration (how to avoid false winners; incrementality caveats; cross-channel reuse policy) - Dependencies on Audience + Landing Page (explicit handoffs) 4) Audience & CRM Integrations Specialist File: outputs/agent_teams_demo/ads_management_automation/13_audience_crm_integrations.md Sections (in this order): - Audience Strategy Map (prospecting, retargeting, lifecycle; table: Segment → Definition → Source → Refresh Cadence → Channel Use) - CRM/Offline Conversions Data Model (fields, IDs, deduping; PII handling; hashing) - Implementation Plan: Meta CAPI + Google Enhanced Conversions (step-by-step; include GTM/server-side options; validation steps) - Lookalikes/Similar + Suppression Logic (who is excluded where; overlap prevention) - Automation & Monitoring (data freshness alerts, match rate thresholds, event volume anomaly detection; table of rules with thresholds) - Privacy/Compliance Guardrails (consent modes, retention, access controls; what to log) - Failure Modes & Mitigations (at least 8 bullet points) - Success Metrics (match rate, event coverage, attributed revenue quality, lead quality feedback loop) 5) Landing Page & CRO Analyst File: outputs/agent_teams_demo/ads_management_automation/14_landing_page_cro_plan.md Sections (in this order): - Funnel Map by Intent (table: Intent → Ad Promise → LP Variant → Primary CTA → Measurement) - Drop-off Diagnosis Checklist (analytics + UX; include page speed, message match, form friction, trust) - Test Backlog (table required) Columns: Test Name | Hypothesis | Page/Step | Variant A | Variant B | Audience/Intent | Primary Metric | Min Sample/Duration | Risk | Dependency Include at least 12 tests across: headline/message match, proof/trust, form UX, offer framing, pricing/plan clarity (if relevant), speed, above-the-fold layout. - Experiment Governance (how to avoid overlapping tests; rollout/rollback; QA) - Tracking Requirements (events, parameters, UTMs, GA4; tie-back to CRM) - Success Metrics & Weekly Review Checklist (10–15 checklist items) - Dependencies on Creative + Audience (explicit handoffs) PHASE 2 — CROSS-AGENT COLLABORATION MECHANICS (REQUIRED) After all Phase 1 files are drafted, run a structured cross-review: - Each specialist must read the other four specialist files (10–14) and add a short “Cross-Review Notes” appendix at the end of their own file (150–250 words) including: - 2 alignments (where another stream supports theirs) - 2 conflicts/risks (where another stream might break or contradict theirs) - 1 question they need answered by the Coordinator Dependency: These appendices must be added before the Coordinator starts Phase 3 synthesis. PHASE 3 — COORDINATOR SYNTHESIS (MUST HAPPEN LAST) Swarm Coordinator produces the unified roadmap, operating cadence, and change log structure. This phase begins only after: - 10_google_ads_automation_plan.md, 11_meta_ads_automation_plan.md, 12_creative_testing_system.md, 13_audience_crm_integrations.md, 14_landing_page_cro_plan.md exist - Each includes its “Cross-Review Notes” appendix Coordinator deliverables (create all files): 1) outputs/agent_teams_demo/ads_management_automation/20_unified_automation_roadmap.md Requirements (1500–2200 words): - Executive Summary (max 180 words) - Unified System Diagram (text-based; describe data flows + decision loops) - Prioritized Roadmap (table required) Columns: Priority (P0/P1/P2) | Initiative | Owner Agent | What Changes | Effort (S/M/L) | Risk (L/M/H) | Expected Impact | Dependencies | Rollback Plan Include at least 15 initiatives across tracking, automations, creative, audiences, CRO, governance. - Operating Cadence (daily/weekly/monthly routines; who does what; meeting agenda templates) - Budget & Experiment Allocation Policy (how to split between channels, tests, and scaling; reallocation triggers) - Risk Register (top 10; each with mitigation and early-warning indicator) 2) outputs/agent_teams_demo/ads_management_automation/21_change_log_template.md Requirements (400–700 words): - Change Log Entry Template (copy/paste format) - Required Metadata (who/what/why/hypothesis/metrics/time window) - Pre-change Checklist + Post-change Validation Checklist - Rollback Criteria (quantitative thresholds + max time-to-rollback) - Audit Trail Guidance (where to store links/screenshots) 3) outputs/agent_teams_demo/ads_management_automation/22_alerts_and_dashboards_spec.md Requirements (900–1400 words): - Dashboard Spec (tables: core KPIs, breakdowns, filters; Google vs Meta vs blended) - Alert Catalog (table required) Columns: Alert Name | Trigger | Data Source | Severity | Owner | Immediate Action | False Positive Risk Include: spend anomalies, conversion drop, CPA spike, ROAS drop, tracking outage, match rate drop, creative fatigue, landing page speed regression. - Data Freshness & QA (what to check daily; acceptable latency; escalation path) FINAL SYNTHESIS/REVIEW STEP (MANDATORY; MUST BE LAST ACTION) After all files are written: - Swarm Coordinator performs a final consistency pass across every file and produces: outputs/agent_teams_demo/ads_management_automation/99_final_review_and_open_questions.md Requirements (600–1000 words): - Consistency Checks Performed (naming, KPIs, guardrails, definitions, conflicts resolved) - Key Decisions Made (5–10 bullets) - Remaining Open Questions (group by Measurement, Creative, Audience/CRM, Bidding/Budget, CRO) - Next 7-Day Implementation Plan (day-by-day checklist) - “If We Do Nothing” Risk Summary (5 bullets) EXECUTION INSTRUCTIONS - Work in the phase order exactly. Enforce dependencies strictly. - Use the Coordinator’s assumptions/definitions verbatim unless you explicitly propose a change; if proposing a change, flag it in Cross-Review Notes and in the final review doc. - Keep recommendations automation-ready: include thresholds, minimum data requirements, and rollback conditions. - Avoid generic advice; provide concrete rules, templates, and tables. - Do not omit the required sections, word counts, or file paths.
Create your own AI agent team at Build Agents Store. Describe your business problem and get specialized agent teams with ready-to-use prompts for Claude Code.