Skip to content
13Symbiont
Close
Symbiont

AI Product Design · Systems · Governance

Symbiont

A Governed Operating System for AI Agents

Role
Creative Director · Product Strategist · AI Workflow Designer
Year
2026
Categories
Product + AI · Innovation · Experience · Brand
Tools
Next.js · TypeScript · AI Agents · Claude · Figma

Overview

Symbiont is an Executive Command Center for a company run with AI agents — a single, governed environment where fifteen registered agents coordinate operations, sales, delivery, finance, quality, and knowledge, all reporting into one operating truth. Every shared action is logged, every external move is gated, and a human always holds final approval.

The Challenge

Autonomous agents are easy to demo and hard to trust. The work was to design an operating model where agents do real work — scouting opportunities, drafting proposals, building estimates, stewarding knowledge — while accountability, pricing authority, and a single source of truth stay firmly human-governed.

Creative Direction

The organizing idea — “Governed agents. One operating truth.” — frames the company as a cooperative network of agents, each with an explicit ID, authority level, parent, and human owner, visualized as a live network graph radiating from a central command. Finance and estimating views carry that same discipline into the numbers: every Rough Order of Magnitude broken into direct labor, burden, operations, contingency, and confidence, with pricing postures (Aggressive / Target / Protect) reasoned out in plain language before a human signs off. The interface language throughout is calm and editorial: monospaced labels, an olive-and-ink palette, oversized numerals, and an always-visible runtime state — shared data, agent authority, external actions.

Process

  1. 01

    Model

    Map the company as a network of governed agents and roles.

  2. 02

    Govern

    Define levels, readiness scoring, and human-approval gates.

  3. 03

    Design

    Build a calm, legible command-center interface language.

  4. 04

    Operate

    Surface a single operating truth across agents, knowledge, and finance.

Selected Views

Agent Network — fifteen governed roles, mapped by authority level and reporting line.
Agent Network — fifteen governed roles, mapped by authority level and reporting line.
Knowledge — “Find the work. Trust the source.” A single, stewarded library of standards, templates, and decisions.
Knowledge — “Find the work. Trust the source.” A single, stewarded library of standards, templates, and decisions.
Finance — a Rough Order of Magnitude estimate, broken into cost lines, contingency, and confidence.
Finance — a Rough Order of Magnitude estimate, broken into cost lines, contingency, and confidence.
Estimating agent broadcast — a pricing posture shared across the network, with exact pricing still requiring human approval.
Estimating agent broadcast — a pricing posture shared across the network, with exact pricing still requiring human approval.

Outcomes

  • Designed a governed operating system where 15 AI agents do real work under human approval.
  • Built a stewarded knowledge library and a transparent, agent-assisted estimating workflow.
  • Created a calm, legible product language for an AI-operated company.

Next Project

14

Station Coffee