top of page

Design Systems + AI at LINQ

An AI-powered design operations practice: making the design system the source of truth for AI agents — including Claude — across a portfolio of six core products.

pexels-pixabay-356043.jpg

Context

​

LINQ is a K-12 school operations SaaS company. It has a portfolio of six core products. My design work and the design system cover all six.

This case study is about the AI-powered design operations practice I built and implemented: the design system as the source of truth for AI agents, including Claude, so product, design, and engineering can move quickly without the experience drifting from our patterns, brand, and accessibility standards.

Role

Lead UX Designer at LINQ since July 2023. UX Designer at LINQ from September 2021 to July 2023. I own the design system and UX standards across the portfolio, keeping patterns, accessibility, and brand consistent across products.

I serve as the cross-functional bridge between product, engineering, QA, and marketing on how components, AI rules, and brand carry into the product. I mentor, and I give directors insights they can use for accurate performance reviews.

Problem

​

Product managers have the design system in Claude and can build rapid proofs of concept and deliverables themselves. That approach significantly helped LINQ deliver during surge funding.

Without an owner for the system and the standards, AI output drifts from our patterns, brand, and accessibility standards. The question the practice is built to answer is how to keep the system, the standards, and the communication consistent across the portfolio, and how to bring these processes to every product team so the same efficiency and consistency show up everywhere.

pexels-picjumbocom-196645.jpg
pexels-startup-stock-photos-212286.jpg

Process

​

I built and implemented a working method around the design system in Claude, product-manager prototyping, UX review, and governance — not a second set of mockups.

The design system in Claude. Product managers have the design system in Claude and build rapid proofs of concept and deliverables themselves.

Edit in the PM build. I review and edit directly in those PM builds rather than making duplicate mockups. This case study does not include before-and-after visuals, because that is not how the work is done.

UX review checkpoints. Work goes through a UX review so component and experience consistency can be checked before it moves forward. Weekly product–UX syncs are used to stay aligned. For complicated pages and bigger lifts, UX designs those directly.

Drift log and governance. I designed a governance model, including a drift log, so speed does not turn into rework. (See Governance, below.)

Accessibility at LINQ. As I design, I do high-level accessibility audits, and I partner with development to ensure compliance across the build.

Solution

​

The solution is an AI-powered design operations model in which the design system is the source of truth for AI agents, including Claude:

The design system lives in Claude and is available to product managers, who generate on-brand proofs of concept and deliverables themselves.

UX runs review checkpoints for component and experience consistency, and designs complex, heavy-lift pages directly.

A governance model, including a drift log, is in place so speed does not turn into rework.

LINQ ERP is the pilot for this model: product managers build proofs of concept with Claude and the design system; UX reviews that work and designs the heavy lifts.

Phased rollout

I designed a four-phase rollout so the practice can be piloted, refined, and then carried to other products — without treating speed as a one-time burst.

Phase 1 — Pilot. LINQ ERP is the pilot. Product managers build proofs of concept with Claude and the design system. UX reviews that work and designs the complicated pages and bigger lifts. During my leave, day-to-day design work, the review checkpoints, and the drift log are covered by the other designer. A retro on my return is meant to show where the guardrails hold and where they break.

Phase 2 — Retro and refine. On my return: reflect on how the pilot went, evaluate gaps and needs, and fine-tune rules, prompts, and process.

Phase 3 — Scale and integrate. Carry proven processes and models to other products, one product at a time, with a repeatable kit and office hours as teams ramp up. Sync engineering’s AI agents with the Claude design system so components, rules, and updates stay aligned. That integration is part of the planned rollout, not work already completed. Timing for this phase is to be agreed after the retro.

Phase 4 — Ongoing. Ongoing oversight, quarterly audits, and a roadmap as AI tools change.

We do not have baseline numbers. Success for the pilot was planned to be judged by signals the work itself produces — the drift log, the weekly syncs, and the retro — not by metrics we do not have. This page does not claim pilot results.

Governance and collaboration

I designed a governance model so speed does not turn into rework. Part of that model is a drift log: a record of what the AI gets wrong, and where the guardrails hold or break. The biggest recurring drift issues the log is designed to target are component states, advanced grid features and functionality, and layout consistency.

The drift log is part of the model I designed; it has not produced results yet. It is not presented here as a completed measurement program.

The review checkpoint is the main guardrail: AI-built work is not meant to move into development without UX review, and the drift log is there to record misses.

As I design, I do high-level accessibility audits, and I partner with development to ensure compliance across the build.

The model also includes syncing engineering’s AI agents with the Claude design system so components, rules, and updates stay aligned. That step is part of the planned rollout — working with engineering to connect development agents — not work already completed.

I act as the bridge across product, engineering, QA, and marketing: product managers build with Claude and the design system; engineering’s agents should turn designs into code that matches the design; testing should follow the same standards; brand guidelines should carry through into the product experience. UX is the source of truth for components, rules, patterns, and brand expression across those four.

Outcome and status

Since product managers started prototyping in Claude, these are the changes we can already describe. They are qualitative. We do not have baseline numbers, and none are claimed here:

Faster concept turnaround

Fewer design rounds

More ideas tested

Product managers can build rapid proofs of concept and deliverables themselves

The approach significantly helped LINQ deliver during surge funding

Processes I established that are still in use include design reviews, and handing the design system to product so teams can turn work around faster.

The LINQ ERP pilot is the test of the model, including during my leave. This page does not claim pilot results. Connecting development agents remains a planned step.

Learnings

The governance model is aimed at the drift that shows up most often in this work: component states, advanced grid features and functionality, and layout consistency. Editing in the PM build, rather than remaking mockups, is how the work is actually done — which is why this study has no before-and-after visuals. A phased rollout — pilot, retro, then scale, then ongoing audits — is how the practice is meant to move from one product into the rest of the portfolio without losing the review checkpoint that keeps speed from becoming rework.

Screenshot 2025-08-19 at 2.26.31 PM.png
bottom of page