How We Built the Team: Six AI Agents, One Operations Stack

Right Aim runs on six AI agents. Operational agents that build products, manage infrastructure, track metrics, and grow the business. Here's how we designed them.

Start With Roles, Not Tools

The biggest mistake in AI operations is starting with the technology. Instead, we started by defining the organizational roles we needed: Aston plans, Bob builds, Leah grows, Metrick measures, Mira operates, Rex sells. Each role has clear boundaries, clear inputs, and clear outputs.

The Communication Pattern

Our agents communicate through structured channels, not free-form chat. Aston produces weekly plans. Bob consumes those plans and produces build artifacts. Leah reads both to create content. This is a directed acyclic graph — information flows in one direction with no circular dependencies.

The Technology Stack

Each agent runs as a collection of n8n workflows on our self-hosted instance. Claude provides the reasoning layer. Supabase stores operational state. The key decision: agents are stateless. All state lives in the database. Any workflow can be restarted without losing progress.

What We'd Do Differently

If we started over, we'd invest more in the observation layer from day one. Metrick was the last agent we built, but should have been the first. You can't improve what you can't measure.