Crafting an AI-Assisted
Design Tool

During my time at SAP's Intelligent Experience team, building AI agent workflows and exploring AI as a design tool, I noticed a critical gap: while everyone was enthusiastically prototyping, nobody had answered a fundamental question. In what part of the design process can AI actually create value? Will AI amplify designers or replace them?

Rather than explore randomly, I chose a systematic approach: the Design Thinking framework. It's rigorous, widely understood, and cyclical — perfect for understanding where AI could genuinely help without replacing human judgment.

My question: at which stages does AI make designers faster, smarter, and more confident, while keeping them in control?

Mapping AI Onto Design Thinking

I mapped the complete Design Thinking process — Empathize, Define, Ideate, Prototype, Test, and Hand-off — and built a Figma diagram to visualize all six phases. For each phase, I identified the key tasks designers actually do, then mapped where AI could genuinely add value, and what role it should play: evaluate, plan, or generate.

From Diagram to Working Prototype

Using Figma's "Send to Figma Make" feature, the diagram became a functional prototype. But to build something more powerful, and to think through the design more deeply, I brought it into Claude Code. Using Figma's MCP integration, I moved between the two tools — designing in Figma, building in Claude Code, testing and refining — until Relay emerged: an AI-assisted design tool that guides designers through Design Thinking phases while keeping them in control.

Relay's dashboard gives designers a home base across projects — resuming a project mid-stage, seeing AI-generated activity at a glance, and getting oriented on how the tool works before diving in.

Documenting the Process: The Design Hub

As I built Relay, I created a Design Hub — documentation capturing my process at each phase, built and refined the same way, moving between Claude Code and the browser. It became a reference guide the team could use to understand the Design Thinking approach and apply it to future projects, making the process transparent and repeatable.

AI Is a Multiplier: Control It, or It Controls You

AI makes fast things faster and hard things approachable, but it doesn't remove the need for judgment, taste, or deep thinking about the problem. The designer's job shifts from execution to direction: knowing what to ask, reviewing everything, and deciding what's actually good.

The challenge is that AI produces a lot of confident-sounding output. Left unchecked, it's easy to accept without scrutiny. What worked:

• Treat AI output as a starting point, not a conclusion.
• Give specific, narrow instructions.
• Ask for three bullet points, not a summary.

Control the output before it controls you. This is where Relay's real value emerged. It's not the tool itself that matters — it's what the tool revealed: AI amplifies human judgment, it doesn't replace it.

Design Leadership Remains About Understanding, Not Tools

Relay proved something important: great design doesn't come from better tools. It comes from understanding users so deeply that solutions feel inevitable. AI can help you work faster, generate more options, and document better, but it cannot do the hard thinking about what users actually need. As AI handles more execution work, design leadership becomes more strategic: judgment, empathy, and the ability to ask the right questions.