All work
beta

AI Marketing Workforce

A multi-agent system that runs marketing like a team

Role: Designer & builder — agent architecture and orchestration

ClaudeMulti-agentNext.js 16SupabaseAI SDK
multi-agent · orchestrated5 specialists
CAMPAIGNANALYTICSCONTENTSEOAUDIENCEORCHESTRATOREXECUTIVE
· The problem

Marketing at scale is a dozen specialist roles — campaign management, analytics, content, SEO, audience research — coordinated by someone who can see the whole board. That coordination is exactly the part that used to need a large team.

· What I built

A multi-agent system built around a Marketing Executive orchestration agent that plans the work and delegates to specialists: a Campaign Optimizer (ROAS, A/B tests, bid and budget reallocation), an Analytics & Reporting agent (cross-channel aggregation, anomaly detection, attribution modelling, trend forecasting), plus content, SEO and audience agents — each owning its slice and reporting back into one view.

· How I built it
  1. 01An orchestrator that decomposes a goal into tasks and routes each to the right specialist agent.
  2. 02Specialist agents with scoped tools and structured outputs, so results are machine-checkable, not just prose.
  3. 03A shared state and reporting layer that rolls specialist results up into a single cross-channel picture.
  4. 04Claude for the reasoning, with deterministic guardrails around anything that would spend budget.
· Outcome
1 + 5
orchestrator + specialists
Cross-channel
unified reporting
Parallel
agent execution
  • Coordinates the work that previously needed a roomful of specialists.
  • Campaign, analytics and content tasks run in parallel and report into one place.
  • Any figures shown inside the demo UI are illustrative sample data.
· What this proves
  • Multi-agent orchestration design (a planner driving specialist agents)
  • LLM tool-use and structured outputs in a real workflow
  • Attribution, anomaly detection and reporting logic
  • Turning an operator's playbook into an automated system
· What's next

Tightening the guardrails around spend and adding more specialist agents as the playbook grows.