OPPRO — AI-ASSISTED PRODUCT DESIGN
Organizing an AI-powered agency to bridge strategy, design, and product development.
At Oppro, I define a working model that combines human leadership with a coordinator and 6 specialized agents in Claude Code. I organize roles, instructions, and context to tackle strategy, design, development, and business tasks.
Through this process, I built three proprietary tools: a UX/UI auditor, a foundations and tokens generator, and a commercial pipeline.
Role
AI-Assisted Product Designer
Company
Oppro
Year
2026
Platforms
Claude Code
The problem
Designing a one-person agency operating model capable of tackling digital projects that require diverse skill sets: strategy, UX/UI, Design Systems, development, and business.
What made it hard
The challenge was to organize artificial intelligence capabilities into a coordinated system, avoiding siloed agents, inconsistent instructions, and difficult-to-maintain workflows.
To solve this, it was necessary to define responsibilities, structure shared context, establish coordination mechanisms, and maintain human supervision over strategic decisions and output quality.
What I decided and why
I designed a structure featuring central coordination and specialized agents, guided by my product and design judgment.
I applied this model to building three web tools: a UX/UI auditor, a foundations and tokens generator, and a sales pipeline.
Strategy & Scope
I defined the agency's focus and the tools to build, aligning business needs with design and automation opportunities.
Agent Organization
I distributed responsibilities between a coordinator and 6 specialized agents to tackle different tasks across the process.
Instructions & Context
I designed prompts, instructions, and roles with shared context in Markdown files to guide agentic workflows in Claude Code.
Design & Build
I led the definition of flows and interfaces as well as the AI-assisted build, refining the tools throughout development.
The result
Specialized Coordination
A coordinator and 6 agents with defined roles and instructions.
Three Web Tools
A UX/UI auditor, a foundations and tokens generator, and a sales pipeline.
A Foundation for Evaluation
I structured Markdown files to define the purpose and instructions for each role, providing context for the work in Claude Code.
How the agents are organised
My role
As a Product Design Lead with hands-on experience and intermediate expertise in AI-assisted design, I designed and structured a one-person agency operating model based on specialized agents within Claude Code.
I organized strategy, product design, marketing, sales, development, operations, and data analytics functions using Markdown files with distinct instructions and responsibilities.
I also defined an overall coordination structure and project documentation, organizing agentic workflows while maintaining a human-in-the-loop approach—keeping strategy, priorities, design decisions, and review of outputs under human oversight.
This project explores how to apply Design Operations and team organization principles to emerging AI-assisted work models.
- Define Oppro's vision and priorities.
- Organize agent roles and responsibilities.
- Establish instructions and working context.
- Lead product, UX/UI, and design system decisions.
- Coordinate the development of proprietary tools using Claude Code.
Proposed Evolution: Agent-Ready Design Systems
The proposed next step connects this practice with my Design Systems background: structuring tokens with explicit values, components with variants and states, and usage guidelines so agents can query the relevant context for each task.
Evolving toward an agentic system would introduce proposed changes subject to human review and defined autonomy limits. This capability is not yet implemented or validated in Oppro.
Reference framework: Agentic Design Systems · Supernova
Key Takeaways
Organizing work with AI requires making decisions explicit that often remain implicit within a team: who does what, with what context, and under whose guidance.
Building Oppro allowed me to explore this organizational structure and apply it to concrete products. The next step is evaluating its usefulness, identifying flaws, and iterating on the process based on evidence.
The tools have not yet undergone user testing or formal QA; measuring usage and actual impact remains a work in progress.

