Designing for AI, Designing with AI, and Leading Design in the AI Era
Why our posture as builders needs to change
For the last five years, I led the AI/ML design team at AWS. During that time, our team supported more than 30 AI/ML services, launched 700+ features, and helped create 13 new services from 0 to 1. Collectively, this portfolio generates billions in annual recurring revenue—Amazon Bedrock accounts for a significant share of that. I recently left AWS, and as I was exploring new opportunities, I found myself reflecting on how AI is changing Design. Not just tools or workflows, but our entire posture as builders.
For decades, we’ve lived in a world of building deterministic, rules-based software. In that world, the rules were clear, and everyone could stay in their lane. Designers mapped user flows and accounted for all the states an end-user might encounter. Researchers ran usability studies to observe behavior, identify pain points, and evaluate how well users could complete tasks. We optimized for the happy path and worked to smooth out the edge cases. Engineers tested in controlled environments, slowly rolled out features to small cohorts of live customers, and measured success before full launch. Design systems reinforced consistency and scale. In short, we believed we had control.
AI breaks that model.
The probabilistic nature of AI means product behavior at runtime can diverge from what was planned before launch. Building products has shifted from creating deterministic, rules-based software to managing probabilistic experiences. Instead of designing for a happy path with a few edge cases, we now face a future where every experience becomes an edge case. In response, Design must shift toward building systems that adapt in real time, personalize for the individual, and remain resilient under stress.
Designing for AI
How interfaces and interaction patterns change when products are driven by inference rather than deterministic logic
AI introduces a new design material, built on a modern software stack that includes vector databases, prompt management systems, model inference, and agents that can reason, adapt, and act. Designers need to understand this stack to connect it meaningfully to the UI layer. These are our new colors and paintbrushes.
Designing for AI demands new interaction models: systems that clarify confidence levels, accept correction, adapt to individual context, and remain resilient even when the model gets things wrong. It’s a world where interface design meets orchestration, and where UX can bridge model behavior and human trust.
That’s why I created DesignPatternsAI.com. It’s designed to help designers build their intuition and vocabulary around AI-native interaction patterns. Patterns are organized into a five-phase framework: AI Onboarding → User Input → AI Output → User Feedback → System Learning. This broad loop contains smaller internal loops, and across these phase you’ll find 15+ patterns such as:
Expressive Input: lightweight, visual, or emotionally rich ways to shape AI behavior beyond text
Inline Correction: letting users adjust tone, length, accuracy, or content without restarting the entire flow
Memory Control: giving users visibility into what the system remembers, and the ability to edit or erase it
These 15 patterns aren’t hypothetical. They’re showing up in real products, and they reflect the types of challenges every designer now faces when working with AI systems. I also regularly draw inspiration from AIverse.Design, a growing collection of ~200+ real-world examples where the design community is pushing the space forward.
My goal with DesignPatternsAI wasn’t just to document UI. It was to spark momentum. To help designers develop AI fluency and build a common language we can use with PMs and engineers as we deliver work.
Designing with AI
New tools and methods expand where and how Design contributes within the traditional product development lifecycle
Roles across product, design, and engineering are converging into something new: the AI builder. Designers are shifting right, using tools like Subframe to design and hand off pixel-perfect code to engineers who implement business logic, rather than referencing Figma. At the same time, tools like Lovable, Bolt, and V0 enable designers (or anyone) to ‘shift left’—borrowing the engineering mindset of moving upstream into earlier phases like ideation, experimentation, and prototyping. The power of “learn by doing” is available to everyone.
The tools will evolve, but one thing is clear: AI is coming for your tasks. Whether it’s marketing, technical writing, frontend development, UX design, visual design, or management—every role will shift, merge, and expand.
The future belongs to those who embrace these tools. Builders who prototype and ship code, build flexible systems for the AI era, and stay curious through the transition. Look for companies where you're empowered, even expected, to disrupt your way of working. This means learning new methods, exploring dead ends, and pushing toward your highest potential.
Design Leadership in an AI Era
How org design and team culture evolve when every project includes an AI component
For decades, design leaders managed around constraints: limited time, headcount, or budget. We’ve spent years making hard tradeoffs, with never enough time, budget, or people to do what we knew was possible.
With AI, we will move from managing scarcity to managing abundance. For example, every new feature will include multiple prototypes, user studies, and design variations for review. Not because teams suddenly have more time, but because the financial and opportunity costs of creating them drop dramatically. It will become economically irresponsible not to explore more ideas, run multiple variants in production, and deliver more personalized experiences.
Coordinating people and processes without overlap will be a major challenge. As complexity becomes a constant, leaders must drive clarity across their organizations, build strong partnerships across increasingly overlapping functions such as Design and Product or Design and Engineering, and implement mechanisms that help them lead beyond their direct line of sight.
Transparency becomes a necessary cultural cornerstone for every business. As AI handles more execution, humans will take on more strategic initiatives. But strategy only works when people have clear visibility into what’s happening across systems, teams, and customer signals.
The ability to propose the right priorities, shape the right ideas, and make the right bets depends on access to real detail and shared context. This is the same context your AI systems need in order to differentiate. Without the unique signals, proprietary data, and internal perspectives from your team and your business, everything you build—your product, your solution, even your company—risks being commoditized by the latest frontier model.
I believe the next decade of my career will be focused on three key areas: designing for AI, designing with AI, and design leadership in an AI era. The Overton window is wide open. This is our moment to rethink what Design delivers, how we work, and why it matters.
In the next post, I’ll look at how we got here as we moved from traditional ML → GenAI → Agentic AI.
Dave, as a designer fascinated by this AI revolution, the “new colors and paintbrushes” analogy really resonated with me. I’ve found that poking into understanding the tech stack has been incredibly helpful to understand where AI is going. We definitely should look beyond traditional design boundaries if we want to embrace AI. Building AI fluency will help us speak a better common language with our engineering peers and advocate for human needs. Your point about designing for uncertainty rather than the traditional “happy path” feels like the fundamental shift we all need to internalize.