The Biggest AI Mistake Design Leaders Are Making Right Now
Most design teams are adding AI the way they added dark mode: as a feature, as a wrapper, as a checkbox to show they’re keeping up.
This is the surface-level trap of +AI, where teams take the existing product, add a chatbot, insert a “Summarize with AI” button, label it intelligent, and hope it drives adoption.
Sometimes it works in the short term. Users get faster answers, support teams deflect more tickets, and marketing has something to demo. But beneath that short-term win, the architecture stays the same. The user still has to work around the system instead of with it. The AI is present but not integrated, useful but not trusted.
This is the ceiling of +AI. It treats models like magic helpers instead of embedded logic.
AI+ [Your Product] – Designing the product so AI is built into its structure, influencing logic, workflows, and decisions from the ground up.
You can see this pattern across the wave of AI retrofits in SaaS: email apps that summarize threads without understanding intent, CRMs that recommend next steps without context, BI tools that generate charts from prompts even when the underlying data is flawed. These are not intelligent systems but narrow, brittle experiences that are often impossible to debug.
But the real damage happens at the infrastructure level. When AI is retrofitted rather than architected, you inherit what I like to call the “genAI latency tax”: the compounding cost of prompt round-trips, context switching, and vector similarity searches that were never designed to feel seamless. The system might deliver the correct answer, but the interaction feels mechanical, disconnected from the user’s actual workflow.
The Alternative: AI+ Design
The opportunity is not just faster answers but more intelligent systems.
AI+ is not a feature or an enhancement but an operating model. In an AI+ system, intelligence is built into the product’s structure where models are not just generating text or images but guiding flows, shaping decision logic, and adapting experiences based on real behavior.
This distinction fundamentally changes the role of design. In this model, design becomes infrastructure that defines how the system behaves when no one is looking, how it handles ambiguity, how it balances autonomy with control, and how it guides users through outcomes rather than just steps.
The value shows up not in the prompt but in the pattern.
Consider what happens when you design root cause discovery as an AI+ experience rather than a +AI chatbot. Instead of asking users to interpret complex metrics in isolation, the system guides them through investigation. Instead of displaying data and stepping back, it offers context and explanation. Instead of building a tool that shows information, it builds one that helps users understand and act.
The intelligence comes from multiple sources. The LLM provides narrative synthesis while structured systems contribute live performance data. The design challenge is aligning these inputs in a way that feels directional rather than speculative, helping users move faster while maintaining the ability to investigate, revise, and deepen inquiry.
Why This Matters Now
AI capabilities are advancing faster than organizational learning. The workforce is demanding AI tools that enhance rather than replace human judgment. Early adopters aren’t just more efficient. They’re operating with fundamentally different capabilities.
But most organizations are approaching this as a UI problem. They’re asking their design teams to “make AI usable” without recognizing that AI systems aren’t just more complex interfaces. They’re reasoning partners that require entirely new approaches to trust, transparency, and collaborative intelligence.
The economic stakes are immediate. Organizations that master AI+ design achieve 2-3x faster AI adoption rates, reduce support costs by 40-60%, and create sustainable competitive moats through superior human-AI collaboration quality.
When I say “design” here, I do not mean visual design or UX tasks but cognitive orchestration: the systematic structuring of how intelligent systems reason, communicate, and adapt. This is not decoration applied after engineering is done but the foundational logic that determines whether AI becomes usable or merely impressive.
Where to Start
The shift from +AI to AI+ starts with a fundamental question: What should this system understand about the user’s intent, not just what should it output?
This means designing behavioral contracts before building interfaces. It means prototyping reasoning patterns, not just response formats. It means treating prompt design as system architecture, not copywriting.
Most importantly, it means recognizing that the most important design decisions in AI+ systems often happen before the first screen renders. They occur in the logic that determines how the system interprets uncertainty, how it escalates risk, and how it maintains trust over time.
This is part of a larger set of ideas I’ll be exploring in more detail soon. The window for establishing AI+ design leadership is narrow, but the opportunity is transformational for organizations ready to lead rather than follow.