UX Design And Artificial Intelligence

🪂 How To Make Your Design System AI-Ready (https://lnkd.in/dtnpy7CM), a practical guide on how to reduce drifts, minimize mistakes, maintain context and improve the quality of AI-generated prototypes — with structured spec files, automated auditing and token layers. Put together by Hardik Pandya from Atlassian. --- 🔹 1. Design Decisions Are Infrastructure AI-generated prototypes often don't deliver consistently decent results because of tiny inconsistencies scattered all across a design system. Often it's decisions made but not documented, hard-coded values never cleaned up, or relying too much on AI making sense of mock-ups or design flows on its own. Unsurprisingly, better AI prototypes come from better data — but also from better human guidance. We shouldn’t assume that AI knows how to choose the right component, and how to design with accessibility in mind. It needs priorities, a clear path on how we make decisions, design principles, examples, do's and don'ts. In fact, we should treat design decisions as infrastructure. That means that every time we make a decision — not just a design decision, but even decision on how actually prioritize our work and how we make decisions around here — it must find a path into the spec file that is then consumed by AI. --- 🔶 2. Three Layers: Spec Files + Token Layer + Audit To ensure quality, we establish design principles, guidelines, rules in a form of “spec files”). It's structured Markdown files that include spacing rules, color choices, component usage guidelines, priorities etc. AI is going to read and reuse that spec file every time it's going to generate a prototype. Because the spec files are text files, it's much more cost-effective, but also much more accurate just because we don't rely on AI recognizing or decoding patterns from mock-ups, but gets specific guidelines instead. In fact, extending code is often a more effective way than generating code from mock-ups. Token layer lists and keeps updated all tokens used throughout the design system. AI always chooses from a closed set of named variables instead of inventing plausible values ad-hoc. An audit script catches what AI gets wrong. It scans the prototype and flags every hard-coded value and flags it if necessary. It can be a regular software doing that, with AI waiting for its feedback to come back. Finally, when a design system ships updates, a sync routine flags which spec files need updating. The goal is to make sure that AI always reads up-to-date, current specs, not the ones written against an outdated version. --- 🔺 3. Examples of AI-Ready Design Systems ⌾ Atlassian: https://lnkd.in/dVsGc3Cp ⌾ Carbon: https://lnkd.in/d4zq4WWb ⌾ CMS Design System: https://lnkd.in/dHHzV3en ⌾ Nordhealth: https://lnkd.in/d8C4j2ZA Yet again, AI can’t magically resolve technical debt or design debt — it needs guidance, decisions, priorities and principles.

AI Prototyping 101: If I had to teach someone how to actually build usable products with AI, this is where I’d start. Here's the step-by-step workflow that feels like magic: — ONE - THE UNIVERSAL AI PROTOTYPING WORKFLOW No matter which tool you’re using — v0, Bolt, Replit, or Lovable — this is the backbone of a solid AI build process: 1. Start with Context AI works way better when it knows what you're working with. Figma files are ideal, they give structure and design language. If you don’t have those, use screenshots of your product. Worst case? A hand-drawn wireframe is still better than nothing. Without visual context, AI makes blind guesses. And you’ll spend more time correcting its “creativity” than building useful stuff. 2. Write a PRD (Yes, Even for AI) A simple .md file with a few bullet points on what you’re building goes a long way. Include: - What the customers want - What the feature does - Key user flows - Must-have functionality You can even ask Claude or GPT to write the first draft. But the better your input, the stronger your first output. 3. Get to Building Now open up your tool of choice. Start with a big-picture command. Then zoom in. Don’t say “Build me a dashboard.” Say: “Build a dashboard with 3 sections: recent activity, user goals, and notifications. Each should have X, Y, and Z.” Also, AI can handle technical stuff. So don’t hold back. Use real terms: auth flow, API call, state logic, it gets it. 4. Iterate Like a Builder, Not a Perfectionist Make one change at a time. Test it fast. Roll it back if it doesn’t work. This isn’t “prompt once and ship.” This is real prototyping. AI is just helping you move 100x faster. — TWO - TOOL-BY-TOOL BREAKDOWN (Complete walkthrough of the tools with screenshots, real examples, and tool setups is linked at the end.) So, let’s talk interfaces here. Here’s what each platform does best: 1. v0 - Figma import is seamless - Template gallery = instant jumpstart - Chat interface bottom left, live preview on right - Exports clean code and deploys fast 2. Bolt - Same vibe as v0, but more technical - Built-in Supabase integration with a terminal access - Deploys to Netlify in one click 3. Replit - This one feels like a real IDE - You get an “AI agent” to plan everything - Built-in chat, live console, multiplayer mode - Ships to a live URL, complete with CDN 4. Lovable - The most design-friendly of the bunch - Visual editing > code editing - Figma support, Supabase, live preview, it’s all there - Great for teams who want to stay out of code — I broke it all down - with screenshots, working examples, and use cases - in this full walkthrough: https://lnkd.in/eJujDhBV — All of these tools are powerful. But none of them matter if you don’t understand the workflow behind how to use them. Once you’ve got that down, you can ship real products in hours, not weeks.

Talk less. Prototype faster. The best teams don’t discuss ideas endlessly; they just build them. But how do you get the right prototype fast enough? Most new product initiatives are not about creating a new product. They're about improving existing ones. In other words, they already have a product, customers, and a design language. The machine is slow, perhaps rusty, but it has worked for ages now. Any attempts to improve the process usually failed or gave barely any noticeable improvement. However, this is where the AI comes in and why I’m genuinely impressed with Reforge Build, which has now been launched in beta! It’s an AI prototyping tool made for product teams, not solo builders. It starts where your product already is and accelerates what comes next. Don't take my word for it, try it yourself: Check out Reforge Build and explore what’s possible with AI that actually understands your product: https://lnkd.in/duh4YC_H But why did it impress me? 1) Looks like your product Upload a screenshot or connect to Figma. Reforge Build instantly matches your real design system: colors, fonts, spacing, everything. No endless cleanup. No imagination is needed when painting a vision of a future successful product to the stakeholders. 2) Understanding the context Add your product data, strategy docs, and customer insights. Build the prototypes using your actual tiers, features, and messaging. This won't be just a rough draft, but something your actual design team could have presented to you after weeks of work. 3) Plans before it generates Instead of vague prompts, you define user needs, metrics, and layout priorities. AI creates a plan before generating, so the first version is already close to your vision. After all, you need a workable prototype, not an AI slop wannabe! 4) Explores options, not just outputs This REALLY left me with my jaw on the floor: Reforge Build generates multiple design directions, compares them side by side, and mixes the best ideas. I can only imagine this is the experience of a Product Manager with multiple design teams ready to work on a single project... 5) Works like a team tool, not a solo hack Comment, remix, reuse templates, so your second iteration takes minutes, not hours. Nobody's perfect, not even your AI teammate, but every teammate gets better with proper feedback! Impressive, isn't it? Would such an AI prototype tool speed up your new feature's go-to-market time? Let me know in the comments! #productmanagement #ai #ux

Breaking New Ground in Sequential Recommendation: LLM2Rec Transforms How We Understand User Behavior The recommendation systems powering our daily digital experiences just got a major upgrade. Researchers from National University of Singapore, University of Science and Technology of China, and Singapore Management University have introduced LLM2Rec, a groundbreaking approach that bridges the gap between semantic understanding and collaborative filtering in sequential recommendation. >> The Technical Innovation Traditional recommendation systems face a fundamental challenge: ID-based embeddings capture collaborative filtering signals but lack generalization, while text-based approaches offer transferability but miss crucial user behavior patterns. LLM2Rec solves this through a sophisticated two-stage training framework. > Under the Hood: How LLM2Rec Works Stage 1: Collaborative Supervised Fine-Tuning (CSFT) The system transforms large language models into recommendation-aware engines by training them on user interaction sequences. Instead of predicting generic next tokens, the LLM learns to predict the next item a user will interact with based on their historical behavior. This process embeds collaborative filtering signals directly into the model's understanding. Stage 2: Item-level Embedding Modeling The researchers perform two critical adaptations: - Bidirectional Attention Reform: Converting the decoder-only LLM architecture to support bidirectional attention, enabling comprehensive contextual understanding - Masked Next Token Prediction: Adapting the model to handle the new attention mechanism - Item-level Contrastive Learning: Shifting from token-level to item-level embeddings while preserving collaborative signals >> Performance Breakthrough The results are impressive across multiple domains. LLM2Rec consistently outperforms existing embedding models on both in-domain and out-of-domain datasets, achieving 15% relative improvement on gaming datasets and maintaining strong performance even on completely unseen platforms like Goodreads. What's particularly noteworthy is the model's efficiency - built on the lightweight Qwen2-0.5B backbone, it delivers superior performance while maintaining practical computational requirements for real-world deployment. >> Why This Matters This research represents a paradigm shift toward universal recommendation systems that can be trained once and deployed across multiple domains. By successfully integrating semantic understanding with collaborative filtering awareness, LLM2Rec opens the door to more robust, generalizable recommendation engines that understand both what items mean and how users actually behave. The implications extend beyond technical improvements - this could fundamentally change how we build recommendation systems that truly understand user intent while maintaining the collaborative intelligence that makes recommendations relevant.

Product managers & designers working with AI face a unique challenge: designing a delightful product experience that cannot fully be predicted. Traditionally, product development followed a linear path. A PM defines the problem, a designer draws the solution, and the software teams code the product. The outcome was largely predictable, and the user experience was consistent. However, with AI, the rules have changed. Non-deterministic ML models introduce uncertainty & chaotic behavior. The same question asked four times produces different outputs. Asking the same question in different ways - even just an extra space in the question - elicits different results. How does one design a product experience in the fog of AI? The answer lies in embracing the unpredictable nature of AI and adapting your design approach. Here are a few strategies to consider: 1. Fast feedback loops : Great machine learning products elicit user feedback passively. Just click on the first result of a Google search and come back to the second one. That’s a great signal for Google to know that the first result is not optimal - without tying a word. 2. Evaluation : before products launch, it’s critical to run the machine learning systems through a battery of tests to understand in the most likely use cases, how the LLM will respond. 3. Over-measurement : It’s unclear what will matter in product experiences today, so measuring as much as possible in the user experience, whether it’s session times, conversation topic analysis, sentiment scores, or other numbers. 4. Couple with deterministic systems : Some startups are using large language models to suggest ideas that are evaluated with deterministic or classic machine learning systems. This design pattern can quash some of the chaotic and non-deterministic nature of LLMs. 5. Smaller models : smaller models that are tuned or optimized for use cases will produce narrower output, controlling the experience. The goal is not to eliminate unpredictability altogether but to design a product that can adapt and learn alongside its users. Just as much as the technology has changed products, our design processes must evolve as well.

🚀 I Stopped Designing Alone. I Started Designing With AI. And honestly? It changed my entire UX process. Over the past few months, I’ve been integrating AI Figma plugins directly into my real-world client projects,not as shortcuts, but as thinking partners. Here’s how I actually use them in real projects 👇 1. UX Pilot: My Rapid Prototyping Engine When I receive a PRD or rough client requirements, I don’t jump straight into polished UI. I prompt UX Pilot to: • Generate quick wireframes • Create possible user flows • Explore multiple layout structures This helps me validate direction in hours instead of days. I never ship AI output directly, I refine it with business logic and user behavior insights. 2. Clueify: My Pre-User-Test Check Before showing designs to stakeholders, I run an AI usability audit. It helps me analyze: • Visual hierarchy • CTA focus • Cognitive overload • Attention flow It’s like doing a “silent usability test” before real users ever see it. 3. Stark: Accessibility Is Not Optional Real-world products serve real people. I use Stark to: • Check contrast ratios • Simulate visual impairments • Ensure WCAG compliance Accessibility isn’t a feature. It’s responsibility. 4. Octopus.do: I Structure Before Screens In large projects (especially SaaS dashboards), structure matters more than UI. Before designing anything, I: • Map the entire sitemap • Validate navigation depth • Align user journeys Because messy structure = messy experience. 5. Magician: Fast Ideation Mode When brainstorming: • Placeholder content • Icon ideas • Micro-interactions • Empty states Magician speeds up exploration so I can focus on strategy. 6. MagiCopy: UX Writing That Converts Good UI means nothing without clear communication. I use it to: • Generate button variations • Test tone (friendly vs professional) • Improve clarity Then I humanize it with brand voice. 7. Uizard: From Sketch to Prototype Sometimes clients send hand-drawn ideas. Instead of rebuilding from scratch: I convert sketches → editable wireframes → interactive prototypes. Faster iteration. Faster validation. 💡 My Personal Approach AI doesn’t replace UX thinking. It accelerates it. In real projects, I follow this rule: - AI for speed. - Human for strategy. - Users for validation. The result? • Faster delivery • Better alignment with stakeholders • More time spent on problem-solving • Less time on repetitive tasks And most importantly, better user experiences. If you’re a designer still afraid AI will replace you… It won’t. But designers who use AI effectively? They will replace those who don’t. Let’s build smarter. 💜 Whats your way of design? Comment below👇 UX Pilot AI Clueify #UXDesign #UIDesign #Figma #AIinDesign #ProductDesign #UXResearch #DesignProcess #Accessibility #SaaSDesign #UserExperience #DesignThinking #Prototyping #UXWriting #FutureOfDesign #designtools #uiux

Will Apps Need to Redesign Their Interfaces to Accommodate AI Agents? AI agents from OpenAI, Perplexity, and others can comfortably navigate textual and structured digital spaces but quickly hit barriers when faced with visually oriented tools like Gamma, Canva, or WordPress. These popular applications were designed specifically for human cognitive styles, relying heavily on visual intuition, recognition of subtle cues, and interactions guided by visual metaphors. As we can see from early tests, an AI agent accessing these tools via a browser faces hurdles. The reason: interfaces designed around human perception and intuition become ambiguous or even indecipherable to a purely logic-driven entity. This poses a nuanced design question: to effectively support AI agents, will software companies need to consider creating specialised, agent-oriented interfaces separate from the human-focused UX? The idea isn’t simply about creating more structured web pages. Rather, it suggests building parallel visual experiences explicitly designed around AI cognition, incorporating clear functional signposting, predictable interactions, and logical progressions that agents can reliably parse. The implications are notable: ➡️ Strategic Differentiation: Platforms offering agent-friendly interfaces might attract companies prioritising automation and seamless AI integration, creating new competitive landscapes. ➡️ UX Complexity: App developers will need to strike a balance. How much complexity can they add before negatively impacting the human experience? Can dual interfaces coexist without excessive overhead? ➡️ Productivity and Innovation: With optimised interfaces, agents could more effectively handle complex workflows, opening up new productivity gains beyond basic task automation. Reflections: 🤔 Will AI-friendly UX design become a new competitive advantage? 🤔 How feasible is it for companies to maintain dual-interface platforms for humans and AI agents? 🤔 Will the cognitive divide between human intuition and AI logic become a central consideration in the next era of software design? I'd be very interested in your thoughts. #AI #UX #ProductDesign #FrictionAdvantage

True Humans + AI work and thinking means humans should participate in the AI's thinking processes. A very interesting new paper proposes a "Collaborative Workshop" approach to extended chain-of-thought processes such as deep research. They base their approach on three principles: Transparency: The agent’s reasoning, file system, and terminal outputs are fully visible in real-time. Symmetrical Control: Humans and AI have equal authority to modify the workspace. A human can edit a code file or a plan document just as easily as the agent can. Role Fluidity: The workflow can seamlessly shift between AI-led (autonomous) and human-led (assisted) modes. Beyond the specifics of the approach outlined in this paper, these principles are excellent starting points for all AI interface design. They do this by externalizing the agent's thinking into a visible "Plan-as-Document" markdown file (TODO.md). Users can hit "Pause," edit the TODO.md file to correct the agent's strategy, and hit "Resume." The agent then reads the updated plan and adjusts immediately. Despite being designed for collaboration, the system proves highly capable autonomously. ResearStudio achieved 74.09% on the GAIA benchmark, outperforming OpenAI’s DeepResearch (67.36%) and other state-of-the-art systems. The paper gives concrete examples of how human participants in the collaborative thinking workflow create better results. "It transforms the agent from an opaque, brittle tool into a resilient, trustworthy partner, providing the essential safeguard needed to deploy autonomous systems on complex, real-world problems." Full code available with the paper. Image created by Nano Banana Pro

Your AI-generated code is probably excluding many people. "a11y" is shorthand for accessibility — building digital products that anyone can use, including people with visual, motor, cognitive, or hearing disabilities. Over 1 billion people worldwide. But lots of existing websites aren't taking them into consideration. In 2025, WebAIM found that 94.8% of the top one million home pages have detectable accessibility failures. Sadly, AI does not fix this. Because AI coding tools learn from existing code on the web. And 95% of that code is already inaccessible. The models are reproducing a broken baseline. A 2025 study from Carnegie Mellon found three problems when developers use AI coding assistants: → AI doesn't give you accessible code by default (if you don't ask, AI won't prioritize it) → AI omits many important a11y attributes → AI doesn't verify compliance. Many a11y flows have to be verified at runtime The result is missing keyboard navigation, broken focus management, ARIA attributes sprinkled in for show but wired up wrong — which is actually worse than no ARIA at all. This isn't about AI being bad. It's about a knowledge gap that AI inherits rather than solves. As AI generates more of our frontend code, inaccessible patterns are scaling faster than ever. Every vibe-coded app shipped without accessibility review is another site that excludes people. If you're building for the web, start with these basics: → Use semantic HTML. A button should be a <button>, not a styled div. → Test with your keyboard. Tab through your page. Can you reach everything? → Use headless UI components like Radix, Ariakit, Base UI, etc., they have a11y features built in. → Run a11y checkers like axe DevTools or WAVE. They catch the low-hanging fruit in seconds. → Don't trust AI output blindly. Review it specifically for accessibility. Accessibility isn't charity, it's quality engineering. It should not be an afterthought.

Test Automation for AI Agents? I didn’t believe it until I read this paper. And it completely changed how I think about testing for AI-native products. Here's what blew my mind ↓ Researchers built something called AgentA/B — an automated A/B testing framework where LLM agents replace human traffic. No users. No analytics delays. Just 1,000+ AI agents navigating real websites like Amazon. They: → Search like humans → Click with intention → Filter and purchase → All on live DOMs using Selenium + structured reasoning These aren’t prompt-based toys. They’re persona-rich, behavior-driven agents making real choices in real time. And the kicker? The results actually matched human behavior. Agents exposed to better UX: → Clicked more → Used filters more → Bought more → Spent more ($60.99 vs $55.14) One stat was even statistically significant. That’s wild for synthetic traffic. Here’s what this means for SDETs & AI Agent devs: → You can now test UX before launch → Catch issues early, without waiting on live traffic → Simulate edge cases and underrepresented users → Reduce cost, risk, and iteration cycles It’s a whole new dimension of test automation. Not just pass/fail. But: Did the agent behave like a real user? AgentA/B shows how AI can augment A/B testing—not replace it. If you're working on AI agents, this paper is 100% worth the read. Let me know and I’ll drop you the link. ♻️ Repost this if you’re curious about where test automation is going next. PS: Want to stay updated on AI x Testing? Scroll to the top. Follow Shrey Shah to never miss a post. #AI #TestAutomation #GenerativeAI #LLM #AIAgents #SDET #LangChain #Selenium #ABTesting #AgentSimulation