AI in product
1.1. Integration: integrating AI in workflows
1.1.1. For UX professionals, this means how to effectively integrate AI into your workflow.
1.2. Growth: designing better AI experiences
1.2.1. Understand how AI makes decisions, what users want and expect from AI, and how to customize AI to personal taste.
Accessibility/guardrails
2.1. Core details
2.1.1. Keyboard and screen reader paths, captions, contrast, cognitive load, explainability for low-confidence states.
2.2. Quality assurance
2.2.1. Privacy, security, provenance, and bias checks with audit trails and consent patterns.
Research and testing
3.1. A/B testing is still huge
3.1.1. Faster output means more testing; define guardrail metrics and success criteria.
3.2. Evaluation
3.2.1. Eval writing: representative prompts, gold outputs, scoring rubrics.
3.2.2. Usability on critical tasks; staged rollouts with allowlists and clear rollback plans.
3.3. Derisking launches and releases
3.3.1. QA checklist for hallucinations, stale data, failure modes, and privacy leaks.
Craft and design taste
4.1. Verbalize your taste
4.1.1. How to prompt and get AI to make what you want (prompt engineering).
4.1.2. Fine tuning when justified with clear acceptance criteria.
4.1.3. Eval writing to measure quality and fit.
4.2. Know how to identify problems and what a good experience is, and why
4.2.1. Use the lenses of usability, equity, enjoyability, and usefulness.
4.3. PM + UX merge
4.3.1. Learn to direct AI and people; communicate across different mediums; keep a single source of truth.
4.4. What AI gives you more time to do
4.4.1. Create delightful experiences: motion design, micro-interactions, precise empty states.
4.4.2. Insert meaning behind everything with restraint in structure.
4.4.3. Refine and present deliverables.
4.4.4. Create consistency.
4.4.5. Explore ideas and implementations.
Afterword
5.1. Many stakeholders have adverse perceptions of AI. Use this to your advantage and validate value beyond saving time.
5.2. It is up to you to excel in taste and creation. Do it before everyone else catches up. Constantly test your taste. Balance intention and make meaning.
Designated Readers: UX designers, UX researchers, Product managers, and Design leaders
Introduction
AI changed UX from screens and aesthetics to systems and outcomes. Our craft now lives at the level of meaning and scale. This creates new responsibility and expands our methods, scope, and strategic position. To move forward effectively as a designer and product leader, we must focus on three core practices of the digital designer.
Product Sense
With the rush to integrate new technology so as not to get left behind, there has been a mindset shift for the past four years to lead with an idea first and fail faster, especially with AI. Capability push is outpacing opportunity pull, and the result is a vision that drifts away from user value. With AI, it can seem efficient to develop technology before validating the problem. This undercuts a user centered foundation. The first task for designers is to frame the product strategically: clarify the jobs to be done, justify the use of AI, and measure the outcomes. Think in terms of product sense from a founder or PM’s perspective. To solve a real human problem, designers must be able to define WHY and HOW AI should be used.
This means challenging the default goal of “saving time,” “reducing workload” to identifying where AI adds uniquely meaningful value to the user’s workflow. This means establishing a holistic vision for what a “good” AI experience is; measured using the four pillars: usable, equitable, enjoyable, and useful.
Guardrails: Trust, Safety, and Accessibility
Second, we must build for trust, accessibility, and guidelines. The core problem is that AI systems are not inherently neutral or safe. Without a UX-driven approach, they will amplify bias and erode user trust. The current, broken experience is unregulated, forcing the user to do the heavy lifting of preventing hallucinations with evaluation frameworks, managing context with chunking and summarizing, improving outputs with retrieval-augmentation (RAG), and governing models for compliance.
This is a fundamental design failure. The responsibility for creating these guardrails must belong to the designer. Asking users to manage prompts, context, and evaluations is like handing them an early car on unmarked roads. Designers must build the lanes, signs, and dashboard, not only hand over a complex machine.
Validation and Research
Third, we must validate and measure with rigor. In a world of probabilistic systems, our design intuition is not enough. Evidence and benchmarking, not just performance, are the foundation of a successful AI product. We must evolve our research methods to test not just the interface, but the quality, relevance, and safety of AI output. This means developing robust evaluation rubrics and frameworks to systematically score AI performance against our quality testing plans. We should be able to make data-driven decisions and UX design and research should withstand scrutiny, with decisions documented and replicable.
Design Craft
Fulfilling these imperatives demands an evolution of our craft. The designers who thrive will be those who can verbalize taste, using clear criteria rather than vibes and feelings, to articulate WHY one AI output is better than another. They will treat prompts as design tools and blueprint components, not simply to understand and utilize AI in their workflow, but also to fully understand WHY there is a need for better UX in AI and HOW they can do that.