How Matrix took shape
A product journey from "this looks useful" to "I can actually use this."
01 — Observe
Interest was high. Activation was low.
Interest was high.
Activation was low.
I saw colleagues intrigued by early AI workflows, but almost everyone stopped short of bringing them into everyday work.
The barrier was not awareness. It was the distance between interest and practical use.
02 — Validate
The friction was practical and conceptual.
The friction was practical
and conceptual.
When I explored why people dropped off, I found the hesitation was both practical and conceptual.
The problem was not just setup. People needed a clear mental model before they could trust reusable AI capability.
Using it felt technical
- Setup interrupted momentum
- Installation/configuration felt unfamiliar
- Ownership and runtime were unclear
Understanding it was equally hard
- AI was often understood mainly as chat
- Skills and agents were not clearly differentiated
- Predictable steps and probabilistic models were conflated
- Autonomy created questions about trust and control
What does it do?
What is predictable? What may vary?
What can it access or change?
When does a person step in?
03 — Define
Reuse before reinvention.
Reuse before
reinvention.
Instead of asking everyone to build from scratch, I framed a simpler bet: make existing capability easy to find, understand, and reuse.
Start from capability that already works. Create new only when necessary.
Useful AI capability existed, but colleagues struggled to: understand it → activate it → trust it
Make reusable AI capability understandable and usable without requiring people to become AI engineers.
Reuse before reinvention.
Colleagues who see useful AI-assisted work but hesitate at technical setup and unfamiliar concepts.
Find, understand and activate reusable capability.
Move from "That looks useful" to "I can use this."
04 — Build
Matrix 1.0 — Make capability usable.
Make capability
usable.
I built Matrix as a lightweight home for reusable skills and agents: clear purpose, visible outputs, ownership, setup, and a practical next action.
Distribution became part of the product, not an afterthought.
A guided form generated complete skill packages and structured review proposals, removing the requirement for manual Git setup.
05 — Test & iterate
What people struggled with changed the product.
What people struggled with
changed the product.
Watching people try the prototype revealed friction that led directly to three core experience changes.
Validation changed the system, not just the screens.
06 — Learn
Responsibility followed adoption.
Responsibility followed
adoption.
Once capabilities entered everyday use, the questions changed from "How do I run this?" to operational trust:
- What is this agent allowed to touch?
- What happened during execution?
- When does a human need to intervene?
Adoption was only the first problem. Operational trust and responsibility came next.
07 — Reflect
Three principles to carry forward.
Three principles to
carry forward.
- Reuse should be easier than reinvention.
- Explain the job before the technology.
- Hide complexity, not responsibility.
The best agent products don't start by explaining agents. They start by delivering human value.
