Point of view · Design systems
The ground layer
A design system should help teams make consistent decisions—in Figma, in code, and as the product changes. A shared foundation, with room for the product to evolve.
A foundation is useful when the next person can understand it, work with it, and keep it up to date.
My approach starts with one product surface. Identify where design and implementation disagree, define the shared tokens and component behavior, and establish checks that help the team maintain them.
The goal is a foundation that designers, developers, and AI tools can work from—with clear rules and room for the product to evolve. That means documenting component states, making shared values easy to find, and deciding which requirements can be checked automatically and which still need review.
Three things the system needs
Shared decisions, a maintained connection to code, and a way to catch inconsistencies.
Shared values
Define the tokens
Give color, typography, spacing, and motion meaningful names. Map those names between design variables and code so the team can see which values belong together.
A maintained connection
Connect design and code
Agree where changes originate, how they reach the other environment, and who reviews them. MCP can be part of that workflow; the integration and ownership need to be defined for the project.
Practical checks
Catch inconsistencies early
Choose checks for the requirements that can be tested, such as token usage or defined component states. Keep human review for interaction quality and decisions that need context.
How the layers fit together
Tokens define shared values. Components give those values behavior and structure. Patterns combine components for recurring tasks, and product screens use them to meet a specific need. The terrain diagram shows how those layers depend on one another.
Make the decisions explicit
This illustrative comparison shows the same kinds of interface decisions described in two ways. On the left, values and component boundaries are ambiguous. On the right, shared names make them easier to discuss, implement, and check.
Example A · unspecified
Example B · defined
Keep design and code connected
Shared token definitions provide a reference between Figma and production code. The team also needs an agreed process for publishing changes, reviewing differences, and resolving conflicts. Where automation is useful, the integration should make those responsibilities explicit.
Start with one surface
A focused engagement makes the work easier to assess. Start with a surface the team uses, agree on what needs to become shared, and document how the team will maintain it. These four steps describe my proposed approach.
Step 01
Compare design and implementation
Choose one surface and identify the differences that matter to its users and the team.
- Compare Figma variables with the values used in production.
- Identify one-off colors, spacing, and typography.
- Agree which differences to address first.
Step 02
Define shared tokens
Start with the values this surface needs and give them names the team understands.
- Name color, typography, spacing, and radius by purpose.
- Map design variables to the project’s code conventions.
- Test how a change is published, reviewed, and applied.
Step 03
Define components and their states
Specify how the components behave, what they contain, and which access requirements they need to meet.
- Prioritize the components used on the selected surface.
- Document states and identify suitable automated checks.
- Review interactions and edge cases with design and engineering.
Step 04
Make maintenance part of the work
Define how the system will change as the product grows, including ownership and review.
- Document where changes start and how they reach design and code.
- Choose useful measures, such as unresolved differences or update time.
- Leave the team with documentation and a process it can maintain.
Related work
These case studies show the product contexts behind this point of view: incident investigation, operational planning, and fleet monitoring. Each describes the interface work; the approach above describes how I would structure a design-systems engagement.
Microsoft 365
Performance Monitoring
Status-aware incident cards, advisories, and impact views for Microsoft 365 admins.
Oil & gas
SBI
An operations dashboard bringing inventory, maintenance, and forecasting information together.
Armada
Atlas — Fleet Map
Fleet tracking, route tracers, and AI-synthesized incidents alongside a drone’s live view.
Talk about your system
Where are design and implementation becoming harder to keep aligned?