User Experience

𝗡𝗮𝗶𝘃𝗲 𝗥𝗔𝗚 𝘄𝗼𝗿𝗸𝘀 𝗶𝗻 𝗮 𝗱𝗲𝗺𝗼. 𝗜𝘁 𝗳𝗮𝗶𝗹𝘀 𝘁𝗵𝗲 𝗺𝗼𝗺𝗲𝗻𝘁 𝗿𝗲𝗮𝗹 𝘂𝘀𝗲𝗿𝘀 𝘀𝗵𝗼𝘄 𝘂𝗽. Embed → retrieve → generate looks clean in a notebook. Real requirements break it: → Questions whose answer is spread across many documents → Industry terms that embeddings get wrong → Bad chunks the pipeline never catches → Answers that live in how things connect, not in any single chunk → PDFs full of tables and images a text-only index cannot read These 5 architectures are how serious teams stay ahead in the agentic AI era: 𝟬𝟭 𝗛𝘆𝗯𝗿𝗶𝗱 𝗥𝗔𝗚 → Dense vectors find meaning. BM25 finds exact words. → Reciprocal Rank Fusion combines both ranked lists. → A safe baseline for almost every team. 𝟬𝟮 𝗚𝗿𝗮𝗽𝗵𝗥𝗔𝗚 → Pull entities and their relationships into a knowledge graph. → Retrieve subgraphs and community summaries, not chunks. → Best when the answer lives in how things connect. 𝟬𝟯 𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗥𝗔𝗚 → A planner agent picks the right tool: vector, web, or SQL. → A reasoner agent keeps trying until the answer is solid. → Retrieval becomes a plan, not a single step. 𝟬𝟰 𝗖𝗼𝗿𝗿𝗲𝗰𝘁𝗶𝘃𝗲 𝗥𝗔𝗚 (𝗖𝗥𝗔𝗚) → Grade every retrieval before you trust it. → Correct → answer. Unclear → rewrite the query. Wrong → search the web. → This is what production RAG actually looks like. 𝟬𝟱 𝗠𝘂𝗹𝘁𝗶𝗺𝗼𝗱𝗮𝗹 𝗥𝗔𝗚 → One embedding model (CLIP, ColPali) for text, images, and tables. → One vector index. One multimodal LLM. → No more separate pipelines for PDFs with charts. I built a runnable example for each of the five patterns. GitHub link in the first comment. The best teams in 2026 do not pick one. They combine them — hybrid retrieval inside an agentic loop, with a corrective grader, over a multimodal index. Naive RAG is a starting point, not a finish line. That is why most enterprise GenAI projects stall at the demo. Which of these five becomes the default RAG stack in the next 18 months — and which stays a specialized tool?

🔎 How To Redesign Complex Navigation: How We Restructured Intercom’s IA (https://lnkd.in/ezbHUYyU), a practical case study on how the Intercom team fixed the maze of features, settings, workflows and navigation labels. Neatly put together by Pranava Tandra. ���� Customers can’t use features they can’t discover. ✅ Simplifying is about bringing order to complexity. ✅ First, map out the flow of customers and their needs. ✅ Study how people navigate and where they get stuck. ✅ Spot recurring friction points that resonate across tasks. 🚫 Don’t group features based on how they are built. ✅ Group features based on how users think and work. ✅ Bring similar things together (e.g. Help, Knowledge). ✅ Establish dedicated hubs for key parts of the product. ✅ Relocate low-priority features to workflows/settings. 🤔 People don’t use products in predictable ways. 🤔 Users often struggle with cryptic icons and labels. ✅ Show labels in a collapsible nav drawer, not on hover. ✅ Use content testing to track if users understand icons. ✅ Allow users to pin/unpin items in their navigation drawer. One of the helpful ways to prioritize sections in navigation is by layering customer journeys on top of each other to identify most frequent areas of use. The busy “hubs” of user interactions typically require faster and easier access across the product. Instead of using AI or designer’s mental model to reorganize navigation, invite users and run a card sorting session with them. People are usually not very good at naming things, but very good at grouping and organizing them. And once you have a new navigation, test and refine it with tree testing. As Pranava writes, real people don’t use products in perfectly predictable ways. They come in with an infinite variety of needs, assumptions, and goals. Our job is to address friction points for their realities — by reducing confusion and maximizing clarity. Good IA work and UX research can do just that. [Useful resources in the comments ↓] #ux #IA

Big day for our show Marketing Against the Grain Posted our 200th episode. Passed 2 million downloads. Passed 3 million YouTube views. Nice work Kieran Flanagan Some lessons from doing a podcast for 200 episodes: 1. Cross Promotion Drives Growth - doing guest spots or promo swaps with other shows or being part of a network that promotes you on other shows increases growth significantly by 30-40 percent. 2. Audio and Video at VERY different - what works on audio RSS in terms of content and format often does work as well on YouTube and vice versa. You need a playbook that incorporates both. 3. Feedback is how you grow - Listener round tables, YouTube comments, emails, engagement data make you better. Look at it and make adjustments each week and you get 10x better as those adjustments compound. 4. Guest need time to settle in - The first 5-10 min with a guest is going to get cut as they are getting comfortable and in the flow. You have to plan for that. 5. Guest bring listeners - As someone comes on your show they do bring their audience. Our guest episodes have higher views and downloads than non guest episodes. This is why so many shows are guest driven.

Stop pasting interview transcripts into ChatGPT and asking for a summary. You’re not getting insights—you’re getting blabla. Here’s how to actually extract signal from qualitative data with AI. A lot of product teams are experimenting with AI for user research. But most are doing it wrong. They dump all their interviews into ChatGPT and ask: “Summarize these for me.” And what do they get back? Walls of text. Generic fluff. A lot of words that say… nothing. This is the classic trap of horizontal analysis: → “Read all 60 survey responses and give me 3 takeaways.” → Sounds smart. Looks clean. → But it washes out the nuance. Here’s a better way: Go vertical. Use AI for vertical analysis, not horizontal. What does that mean? Instead of compressing across all your data… Zoom into each individual response—deeper than you usually could afford to. One by one. Yes, really. Here’s a tactical playbook: Take each interview transcript or survey response, and feed it into AI with a structured template. Example: “Analyze this response using the following dimensions: • Sentiment (1–5) • Pain level (1–5) • Excitement about solution (1–5) • Provide 3 direct quotes that justify each score.” Now repeat for each data point. You’ll end up with a stack of structured insights you can actually compare. And best of all—those quotes let you go straight back to the raw user voice when needed. AI becomes your assistant, not your editor. The real value of AI in discovery isn’t in writing summaries. It’s in enabling depth at scale. With this vertical approach, you get: ✅ Faster analysis ✅ Clearer signals ✅ Richer context ✅ Traceable quotes back to the user You’re not guessing. You’re pattern matching across structured, consistent reads. ⸻ Are you still using AI for summaries? Try this vertical method on your next batch of interviews—and tell me how it goes. 👇 Drop your favorite prompt so we can learn from each othr.

Can we talk about touch screen access for a minute? As a blind person, it’s the thing excluding me the most right now. Want to pay for a meal? Of course, sir. Here’s a touch screen card reader. Want a coffee with your hotel breakfast? Sure! Use this touch screen machine. I’d love a new air fryer. Here are our best models - they’re all touch screen. I’ll join this video call from a conference room. Ok, you can log into the call via this touch screen panel. Oh and don’t forget to check-into the room via the touch screen entry system, or you’ll lose the reservation after 10-minutes. Let me just jump up to the third floor. Please choose that floor on the touch screen, then we’ll visually display what elevator you should get in. I’ll jump on the exercise bike. Amazing! Set the speed and distance via the touch screen, please. I really fancy a burger. Great choice. Just order on the touch screen here, please. I could go on. For hours. Literally. Touch screens are such a big part of daily life now, but they’re excluding me as a blind person at every point. Some of them have accessibility built in - screen readers are available - but you need to either train staff to know how to use it from the outset, or you need to switch it on via the touch screen settings (can anyone say ‘ironic?!’). And that’s the thing… Accessibility isn’t about adding an option somewhere deep in a settings menu. That’s a tickbox. It’s about designing things that work for everyone from the start. The more everyday things move to touch screens - card readers, vending machines, gym equipment, even lifts a the more independence I lose. Not because I can’t use technology, but because the technology isn’t designed to include me. This isn’t about inconvenience. It’s about exclusion, and it’s about basic rights of privacy. I need to rely on strangers for things that should be simple and private. I’m being reminded every single day that accessibility is still an afterthought. And I’m being told that my right to privacy doesn’t matter. If we want a world that truly works for everyone, we have to stop designing for the majority and start designing for reality - a reality that includes disabled people too. Otherwise we’re going be included less and excluded more. #DisabilityInclusion #Disability #DisabilityEmployment #Adjustments #DiversityAndInclusion #Content

Candidates overcomplicate Portfolios. Listen, if you’re a Senior Product Designer Avoid: 🔻 6 case studies about absolutely irrelevant products 🔻 3-5 personas from your user research for each study 🔻 5 iteration cycles are explained and shown in granular details 🔻 10+ visuals of wireframes of all fidelities 🔻 Different structure for each case study  🔻 End without an end. We launched… that's it Instead: 💚 2-3 case studies that are relevant to your future employers  💚 Strong problem statement - can be longer than 1 sentence, just has to be crystal clear 💚 UXR methodology and 2-3 KEY insights. 💚 The process with 1 visual + one key insight of your testing 💚 Final solution - 3 visuals max and tell me what I see 💚 IMPACT of your solution. 💚 Your reflections. 💚 Next steps. No one has 15 minutes to spend trying to dig out relevant information. It's your job to think about the UX of your portfolios. Show what matters. Show a skeleton. Show to create an impression, but leave them wanting more. Focus on this for the first 5 applications. Get feedback. Build from there.

Since you're viewing this on social media, this post is for you. It's for everyone who posts on social media. Many folks — including diversity, equity, and inclusion (DEI) leaders and those in the disability and accessibility space — don't realize how easy it is to make social media content accessible. Here's a reference guide to help you remember. Pick one thing. Start there. Make it a habit. Once you do, move on to the next thing. I suggest starting with hashtags. Super easy. I'm always learning and evolving my accessibility efforts. I used to be a little more freewheeling about using animated GIFs and emojis. Animated GIFs can be a problem for me. Now, I only post them on social media platforms that don't automatically play them. (Facebook and Twitter.) Another thing I used to do was add an opening line to arouse curiosity. Then, I'd enter a few blank lines after that. Thankfully, I quickly learned this is a problem for some disabilities where a lot of scrolling is a challenge. Here are the dos and don'ts in the image. (If you need an image with larger text or different color background, please reach out. A black or white background doesn't work for some folks. So, light purple it is.) - Skip the fancy font generator. Don't use fancy font generators or unreadable fancy fonts. - Be thoughtful about using emojis. Don't overuse emojis. - Be descriptive yet concise in describing images (alt text). Don't fill alt text with useless keywords. SEO. Blah. Blah. Blah. - Describe images in alt text. Don't use automatic alt text. - Limited use of animated GIFs. Don't use them in signatures, avatars, or comments. The motion is sickening for some folks. Add alt text, too. - Use one blank line between paragraphs. Don't use a lot of blank lines or omit blank lines between paragraphs. - Use sentence case. Don't use all caps because it has no visual shape. Please stop yelling. - Capitalize the first letter of each word in hashtags and user names. #CamelCaseRocks. The other hashtag is #YouCannotReadLowercaseHashtags in all lowercase - Check the color contrast on your images and text/background with a free tool. Don't assume an image or text/background pairing is accessible for everyone. - Caption your videos accurately. Don't let autocraptions do all the work. Edit them. - Provide transcripts for podcasts and videos with short paragraphs and speaker identification. Don't publish transcripts with big blocks of text. I hope this image is an improvement based on feedback. I changed the font. I change the light grey to light purple. P.S. What will you do differently after reading this? Comment below. Want training on how to make your content accessible? ✉️ me 🔔 Tap the profile bell now 👉 Find more #MerylMots content #SocialMedia #DigitalMarketing #Accessibility Image: Dos and don'ts for accessible social media with the left side listing what to do and the right side showing what not to do to make social media content accessible.

You don’t need better UX. You need fewer decisions 🚀 Founders love to say their product needs “better UX.” So they bring in designers to polish onboarding, clean up the UI, or add clever microcopy. But the core problem remains the same: using the product still feels like work. If you look at most products today, they feel like decision trees (especially SaaS tools). They’re packed with small choices: → Which plan should I pick? → Do I need this setting? → What happens if I click that? And while each little decision seems harmless on its own, they add up and the product starts feeling more like a quiz than a tool. One of my biggest hacks as a designer was to learn that you need to design for the laziest, most impatient user possible. If it works for them, it’ll work for anyone. The old playbook said: “Make decisions easier.” The new one says: Only present decisions when they matter. Default the rest. Automate what you can. The goal isn’t to eliminate choice but reduce the ones that don’t add value. Give users control when it matters, not when they’re just trying to get something done. Good UX isn’t about more clarity. It’s about less cognitive load. The moment you start realizing that, everything changes. That’s also how we try to design at Lovable. When users build with our AI, we try to handle as much as possible in the background. While most tools expect you to set up everything manually, we flip the model. You describe what you want & we handle the rest for you. Based on pure intend. → Want a dashboard? Say it. → Need Google login + Supabase? Say it. → Ready to publish? Just say it. You chat & Lovable builds. And it’s not just UI, we handle logic, backend, and integrations too. Think about products that feel effortless in your day to day. They constantly make smart choices for you: → Apple Pay defaults to your most-used card → Notion AI suggests starting points → Linear pre-fills fields based on context That’s not just “good UX.” It's context-aware UX. OpenAI’s new agent is another perfect example. You don’t click through menus or tweak settings. You just say: “Book me a flight to Berlin next Thursday.” And it’s done. People don’t want more options. They want faster outcomes. They want control without the effort. So here's my take: Great UX means reducing decisions. Hiding complexity and making the product feel like it already knows what the user wants. Because the less someone has to think, the faster they can move. And the best products in the world don't just pretend to look smart, but act smart. If your product still feels like work, it’s not a UX problem. It’s a decision problem. So design for fewer choices and smarter defaults. Super excited to hear your take. Comment below!

Meta is repeating its biggest privacy mistake. But this time, it’s wearable. Five years ago, Facebook shut down face recognition on its platform. It paid billions over facial data lawsuits. Now Meta plans to put it on your face. Inside its Ray-Ban smart glasses. The idea? You look at someone. The AI assistant identify them. Meta even acknowledged internally that the timing is politically convenient, launching it when critics are distracted. That should concern you, because this represents a shift in public anonymity. Facial recognition on glasses changes the default setting of society. From: You control when someone knows who you are. To: Anyone wearing smart glasses might know instantly. But that is not the main problem. Let me explain why: 1) Facial recognition technology is biased. Multiple studies have shown: - Higher false positives for women - Higher error rates for Black and Asian faces - Misidentification risks for Native Americans In policing contexts, this has already led to wrongful arrests. Now imagine: - Being misidentified at a protest - Being falsely accused to be in a crime scene - Being wrongly tagged at secret meeting Bias in facial recognition target the most vulnerable. And the people most affected are those already over-surveilled. 2) Smart glasses can be hacked Unlike other data, facial features are permanent. If the glasses that identified you are hacked, your biometric features will be forever associated with your name. Even if you never put a single photo online. And there is no remedy for this. Because biometric data is not like a password. You can’t reset your face. This creates long-term, irreversible identity theft risks. 3) This is a governance pattern. Meta paid billions in privacy settlements over facial data collection. It previously shut down facial recognition citing privacy concerns. Now: - Internal privacy review processes are reportedly relaxed - Risk teams have less influence - Product speed is prioritized The technology didn’t suddenly become safer. The risk tolerance simply increased. For blind users, recognition could be empowering. For memory loss, it could be convenient. But governance is not about best-case scenarios. It’s measured on the worst-case risks. Who can opt out? How is consent enforced? What happens when governments demand access? How long is biometric data stored? Who audits accuracy and bias? What security measures are implemented? Facial recognition in smart glasses doesn’t just change a product. It changes the power balance between individuals in public space. And once anonymity is gone, it doesn’t return. AI harm is rarely accidental. It is foreseeable. It is debated internally. And then it is deployed anyway. --- 🔔 Follow me, Chiara Gallese, Ph.D., for analysis on AI governance, systemic risk, and why most tech decisions are governance decisions first.