Understanding User Experience

This week Microsoft radically reshaped the #future of the #internet announcing #NLWeb, an open-source framework that lets websites talk directly to #AI #agents. Here's what i think this means: ▶️ Instead of sites being for humans to read & click, sites will increasingly be for AI agents to understand & act on. ▶️ When AI agents not humans become the primary visitors to websites, users won’t search, browse, compare or buy manually. You'll just ask your agent (will be native in ChatGPT, Gemini, Copilot etc) to plan a trip, find a product, or get an answer. That agent will traverse the web to get the job done. ▶️ This is #transformative. For 20+ years, the open web has been monetized through #ads, the whole system depends on #traffic, humans visiting websites ▶️ But Agents don’t click on banners. They don’t watch pre-rolls. They don’t opt in to cookies ▶️ Once sites are for agents not humans, then the economics of the web will shift from attention to action ▶️ This is already happening folks: MSFT demo'd that TripAdvisor, Shopify, Hearst, & Eventbrite are testing NLWeb ▶️ Meanwhile #ChatGPT and #Gemini are evolving into agents that can act across the web ▶️ Stablecoins and microtransaction rails (like Stripe + Bridge) are laying the foundation for a new monetization layer, where agents pay per task (#PPT), not per view or per click. 👁️ Time to lift our heads up folks & see the #BigPicture, which is NOT today's chats in browsers or endless posts speculating on the future of Google Search in an AI first world. >> Pay close attention to this coming Open Agentic Web. #AI #MicrosoftBuild #NLWeb #AgenticWeb #AdTech #Marketing #OpenWeb #DigitalTransformation #Marketers

Microsoft just introduced NLWeb, an open-source project designed to bring natural language interfaces directly to websites. The vision is simple and powerful: turn any site into an AI-powered app that can answer user questions in plain English. This is a clever way to add AI search and Microsoft’s attempt to redefine the web’s architecture for the age of AI. NLWeb is model-agnostic, runs on all major systems, and connects to any vector database. Publishers control their data, tools, and user experience. For geeks: NLWeb builds on familiar open web standards. It lets websites expose their content using formats like schema.org and RSS, which are then indexed and queried using large language models. Each implementation becomes a Model Context Protocol (MCP) server, making the content discoverable by AI agents if the publisher opts in. Initial adopters include TripAdvisor, Shopify, Eventbrite, Hearst, and O’Reilly. Use cases range from restaurant discovery to media recommendations to e-commerce product searches, all powered by conversational interfaces embedded directly into websites. Microsoft hopes NLWeb will do for the intelligent web what HTML did for the document web. If successful, it could mark the start of a decentralized, agent-ready web where websites speak for themselves. That said, NLWeb’s success hinges on widespread adoption by developers who are already overwhelmed by competing standards, privacy concerns, and limited resources. Without a clear monetization path or killer use case, NLWeb risks becoming just another well-intentioned protocol that never escapes the GitHub demo stage.

Some user groups have distinct usability needs, and to design experiences that truly meet those needs, we need to identify patterns in how different users interact with a product. Clustering helps group users based on shared behaviors rather than broad assumptions, allowing UX researchers to uncover deeper insights, optimize design decisions, and improve the overall experience. One of the most common clustering methods is k-means, which groups users around central points based on similarity. It is widely used for segmenting personas and analyzing behavioral trends but requires predefining the number of clusters, which can be a limitation. Hierarchical clustering offers an alternative by building a tree-like structure that reveals relationships between different user groups. This method is particularly useful for mapping engagement levels and understanding how different users interact with an interface. Density-based clustering, such as DBSCAN, identifies areas of high user activity while automatically separating outliers. This method works well for analyzing drop-offs, onboarding friction, and engagement patterns without assuming a fixed number of clusters. Gaussian Mixture Models take a probabilistic approach, allowing users to belong to multiple clusters at once. This is particularly useful for analyzing hybrid user behaviors, such as those who switch between casual and expert usage depending on the context. Fuzzy clustering is another approach that enables users to be part of multiple groups simultaneously. This is helpful when behavior is fluid and does not fit neatly into distinct categories. It is often used in personalization systems where engagement modes shift dynamically. Constraint-based clustering applies predefined business rules to the process, making it ideal for segmenting users based on factors like subscription tiers or access levels. Grid-based clustering, including the BIRCH algorithm, is particularly useful when working with large-scale datasets. Unlike other methods, BIRCH processes large amounts of data efficiently, making it a valuable tool for analyzing heatmaps, session recordings, and high-volume engagement metrics.

Remember when the biggest flex was getting that little USB dongle to actually connect so we could use the internet at home? The struggle was real. Now we’ve got AI filling out forms and navigating websites for us. What a time to be alive! Every time you think you're caught up, something new drops in the AI space. As part of my usual morning routine, I'd opened Dot for a quick read and ended up spiraling into a full rabbit hole exploring NLWeb, MCP, and Project Mariner. Microsoft’s new open project NLWeb is designed to make websites natively compatible with natural language, turning them into smart, conversational interfaces that work for humans and AI agents alike. It’s open, model-agnostic, and reminds me of how HTML once democratized the internet. NLWeb might just do the same for the agentic web. And then there was Project Mariner. An experimental Chrome extension that gives AI agents "hands" inside your browser. It doesn’t ask a site for data via APIs. Instead, it acts like a human: clicking, scrolling, filling forms, all guided by AI (Gemini, in this case), step by step. It felt surreal because just a few months ago in a product interview, I’d said one of the AI enhancements I could see coming for Excel was the ability to perform multi-step actions via prompts. And here it is, alive in Mariner, its clicking buttons, manipulating sheets, executing full workflows based on a single instruction. From querying websites in natural language to AI agents navigating the internet for us, this feels like a foundational shift. The web is slowly becoming navigable not just by humans, but by intelligent agents too. That changes everything. Dot did what it does best. Explained the whole thing without screaming ‘the robots are coming for your job'! Amazing primer even if you're not a techie (actually - specially if you're not a techie), link in the first comment! Don't forget to visit the 'Dig deeper' section at the end! #AI #AgenticAI #ProductManagement #NLWeb #ProjectMariner #MCP #ConversationalWeb #LLMs #Microsoft #Google

The payments gap between Web2 and Web3 is getting very narrow, but we're not completely there yet. I’ve spent years in both worlds, and I can tell you this: Web3 users just want the same experience they get with Web2. Someone buys something on Amazon, they tap a button, the payment happens, and they're done. They don't think about the payment processor, the settlement layer, or whatever bank is involved. It's invisible. I built Coinflow after seeing how complex crypto payments were scaring away normal users. Our fantasy sports app had great features, but onboarding killed us - making people fund wallets, understand gas fees, etc. That was a huge pain point. Traditional payment networks like Visa or ACH take 2-5 days to settle funds. Meaning when an Uber driver gets paid by a customer, they're waiting days for that money. But with stablecoins we can settle payments instantly. The key is, users don't need to know this is happening. Reddit onboarded millions of users to NFTs by never using the word "NFT" - they called them "digital collectibles" and handled all the blockchain stuff behind the scenes. NBA Top Shot let people buy digital moments with credit cards, not crypto. The wallet was just... a wallet. This is what I've learned building payment infrastructure: 1. End users care about speed and simplicity 2. Instant settlement is the killer feature, the native token isn’t very important  3. The best Web3 UX is literally just Web2 UX When merchants can receive money instantly rather than waiting days, it's game-changing for their business. We've seen clients in marketplaces and games increase conversion rates by 30%+ just by making blockchain invisible. If you're building in Web3, focus on what blockchain actually solves (settlement, security, cross-border payments) but present it in Web2 clothing. Our best integrations are ones where the user can't tell they're using blockchain at all. The future is about better payment infrastructure that looks exactly like something people already understand.

80% of AI shopping suggestions come from affiliates—not your site. 😞 Microsoft new MCP standard, NLWeb, will transform AI search as we know it. NLWeb creates a universal way for AI agents to interact with your brand's data via MCP, putting the brand back in the driver's seat. By exposing a "mystore.com/ask" endpoint, AI agents will be able to quickly access: - Product catalog, pricing, availability, offers - Educational content & FAQ's - Customer reviews and social proof - Brand story and differentiation points No more web scrapping. No more outdated information. This matters for a few reasons: 📝 Brands control their narratives, not some 3rd party affiliate site or competitor comparison pages 🎯 Accurate present your product info, differentiation, and offers 🤑 Craft AI-specific offers to boost conversion Given Microsoft's relationship with OpenAI, I wouldn't be surprised for ChatGPT to start adopting it this year. Early adopters already include Shopify, Tripadvisor, and Eventbrite—signaling this technology is moving toward industry consensus. The possibilities are infinite as this standard evolves. Aside from information retrieval, agents will have more autonomy to perform actions like email sign-up, loyalty program enrollment, and direct transactions.

𝗪𝗵𝗮𝘁 1,038 𝘂𝘀𝗲𝗿𝘀 𝗿𝗲𝘃𝗲𝗮𝗹𝗲𝗱 𝗮𝗯𝗼𝘂𝘁 𝘄𝗮𝗹𝗹𝗲𝘁𝘀, 𝘁𝗿𝘂𝘀𝘁, 𝗮𝗻𝗱 𝗨𝗫 𝗶𝗻 2025. 𝗟𝗲𝘀𝘀𝗼𝗻𝘀 𝗳𝗼𝗿 𝗯𝘂𝗶𝗹𝗱𝗲𝗿𝘀 𝗮𝗻𝗱 𝗲𝘅𝗲𝗰𝘂𝘁𝗶𝘃𝗲𝘀 𝗱𝗿𝗶𝘃𝗶𝗻𝗴 𝗪𝗲𝗯3 𝗮𝗱𝗼𝗽𝘁𝗶𝗼𝗻. Only 12% of users say payments are their favorite use case yet 54% use crypto for it. What’s holding back mass adoption despite rising sentiment and evolving infrastructure? The 2025 State of Onchain UX report, conducted by Reown in partnership with Nansen and YouGov, is one of the most detailed explorations of user experience in Web3 today. Based on over 1,038 participants across the US and UK, the report dives into sentiment, wallets, chain behavior, and the disconnect between belief and behavior. 𝗞𝗲𝘆 𝗟𝗲𝘀𝘀𝗼𝗻𝘀: 1. 𝗨𝘀𝗲𝗿 𝗰𝗼𝗻𝗳𝗶𝗱𝗲𝗻𝗰𝗲 𝗶𝘀 𝗰𝗹𝗶𝗺𝗯𝗶𝗻𝗴: 69% now feel secure using onchain apps, compared to 50.5% in 2024. 2. 𝗜𝗻𝘁𝗲𝗿𝗼𝗽𝗲𝗿𝗮𝗯𝗶𝗹𝗶𝘁𝘆 𝗶𝘀 𝗮 𝗽𝗮𝗶𝗻 𝗽𝗼𝗶𝗻𝘁: 48% use multiple wallets just to interact across chains. 3. 𝗕𝗲𝗹𝗶𝗲𝗳 ≠ 𝗕𝗲𝗵𝗮𝘃𝗶𝗼𝗿: Users believe in payments and social apps, but trading remains dominant. 𝗔𝗰𝘁𝗶𝗼𝗻𝗮𝗯𝗹𝗲 𝗦𝘁𝗲𝗽𝘀: 1. 𝗘𝗱𝘂𝗰𝗮𝘁𝗲 𝘂𝘀𝗲𝗿𝘀 𝗼𝗻 𝘀𝗺𝗮𝗿𝘁 𝗮𝗰𝗰𝗼𝘂𝗻𝘁𝘀. Over 50% still don’t understand them. 2. 𝗦𝗶𝗺𝗽𝗹𝗶𝗳𝘆 𝘄𝗮𝗹𝗹𝗲𝘁 𝗲𝘅𝗽𝗲𝗿𝗶𝗲𝗻𝗰𝗲. Support specialized wallets for daily vs. long-term use. 3. 𝗧𝗮𝗿𝗴𝗲𝘁 𝗰𝗵𝗮𝗶𝗻 𝘀𝗽𝗲𝗰𝗶𝗮𝗹𝗶𝘇𝗮𝘁𝗶𝗼𝗻. Ethereum for settlement, Solana for trading, Base for innovation. 4. 𝗖𝗹𝗼𝘀𝗲 𝘁𝗵𝗲 𝘀𝗲𝗻𝘁𝗶𝗺𝗲𝗻𝘁 𝗴𝗮𝗽. Ensure product design matches emerging user intent. This is more than a UX report. Tt’s a strategic compass for anyone building or funding the future of onchain utility. 𝗔 𝗮 𝗰𝗼𝗻𝗰𝗹𝘂𝘀𝗶𝗼𝗻: Mass adoption will depend less on speculation and more on reducing friction, increasing trust, and aligning behavior with purpose. This report provides both the insight and roadmap to get there. Are you building based on real user data or assumptions? What’s the single most important fix that would convert you from a user to an advocate? Which Web3 wallet or app surprised you most this year? The State of Onchain UX – 2025 Produced by: Reown In collaboration with: Nansen and YouGov Lead Foreword: Jess Houlgrave, CEO of Reown Contributors: Hart Lambur(Co-founder & CEO, Across) Malcolm Levy (Founder, RefractionDAO) Ilan Hazan (Co-founder & COO, DASTAN) Sarah Blommaert-Kassem Luc Falempin Katarzyna Majchrzak, Business and Operations Director, MBA, lawyer Francisco Trujillo Enikő Koklács Diane Chan Joele Novelli Brian Smith Emmanuel Evarist André Mendes William Rae Aaron Dodd Michał Moneta, PhD Greg Cignarella Rodri Fernández Touza Greg Cignarella Nejc Žnidar Lorenzo Ceppaluni Lukas Fiedler Eric Forgy Lou Grande Ethan Luc Ido Ben-Natan Mathieu Cottin Tommaso Cervellati, MBA, CAIA Niamkey Kouamé Anton Kolot Akim Benchiha Nazim Morera Michael Aprossine Roberto P. Rahul Yadav⚡️

Numbers tell you what happened. They never tell you why. This is the biggest blind spot in digital optimization today. Your analytics show where users abandon your digital experience. But the real reason they leave is almost never what your data suggests. Your bounce rate shows people leaving your product page, but it doesn't reveal the confusion they felt when comparing options. Your funnel analysis identifies drop-offs but misses the anxiety triggered when your shipping information appeared after they entered payment details. After optimizing digital experiences for companies like Adobe and Nike for over 16 years, I've seen this disconnect repeatedly. It occurs because of two powerful psychological forces: 1️⃣ Confirmation bias leads your team to interpret data in ways that confirm existing beliefs. "Customers want more features" becomes the lens through which all behavior is filtered. 2️⃣ The availability heuristic causes users to make decisions based on information that's readily accessible... not necessarily what's most important. I witnessed this firsthand with a client who spent months optimizing their product pages based on heatmaps and click data. Conversions barely moved. When we finally conducted qualitative research, we discovered users weren't leaving because they disliked the product... they simply couldn't tell which of the seven (!) options was right for their specific need. The solution wasn't in the quantitative data. It was in understanding the psychological barriers their analytics couldn't capture. The most powerful optimization approach combines: ↳ Analytics to identify WHAT is happening ↳ User research to understand WHY it's happening ↳ Psychological principles to determine HOW to fix it Are you listening to what your data is saying... or what it's hiding?

Look at what they do, not just what they say. User behavior is how users interact with and use software. It includes things like: → how people navigate the interface → which features people use most often → the order in which people perform tasks → how much time people spend on activities → how people react to prompts or feedback Product managers and designers must understand these behaviors. Analyzing user behavior can enhance the user experience, simplify processes, spot issues, and make the software more effective. Discovering the "why" behind user actions is the key to creating great software. In many of my sales discussions with teams, I notice that most rely too heavily on interviews to understand user problems. While interviews are a good starting point, they only cover half of the picture. What’s the benefit of going beyond interviews? → See actual user behavior, not just reported actions → Gain insights into unspoken needs in natural settings → Minimize behavior changes by observing discreetly → Capture genuine interactions for better data → Document detailed behaviors and interactions → Understand the full user journey and hidden pain points → Discover issues and opportunities users miss → Identify outside impacts on user behavior Most people don't think in a hyper-rational way—they're just trying to fit in. That's why when we built Helio, we included task-based activities to learn from users' actions and then provided follow-up questions about their thoughts and feelings. User behaviors aren't always rational. Several factors contribute to this: Cognitive Biases ↳ Users rely on mental shortcuts, often sticking to familiar but inefficient methods. Emotional Influence ↳ Emotions like stress or frustration can lead to hasty or illogical decisions. Habits and Routine ↳ Established habits may cause users to overlook better options or new features. Lack of Understanding ↳ Users may make choices based on limited knowledge, leading to seemingly irrational actions. Contextual Factors ↳ External factors like time pressure or distractions can impact user behavior. Social Influence ↳ Peer pressure or the desire to conform can also drive irrational choices. Observing user behavior, especially in large sample sizes, helps designers see how people naturally use products. This method gives a clearer and more accurate view of user behavior, uncovering hidden needs and issues that might not surface in interviews. #productdesign #productdiscovery #userresearch #uxresearch