Integrating Customer Experience Tech

Accor just made a move that should scare every hotelier who still thinks “AI = chatbot on the website”. Accor launched the ALL Accor experience inside ChatGPT, so travelers can search hotels in natural language, see public + loyalty-member rates, and then get redirected to Accor’s booking flow to complete the reservation. Alix Boulnois, Accor’s Chief Business, Digital & Tech Officer, framed this as a “pivotal moment” for how people will discover and book travel. That’s not PR fluff. That’s a distribution strategy. Because the real story isn’t “wow, cool feature”… it’s this: The hotel website is no longer the start of the journey The start is becoming: one prompt. And the hotels that win will be the ones that can answer instantly: “I’m landing at 07:30, can I early check-in?” “Spa slot tomorrow at 17:00?” “Airport transfer + baby cot + late checkout?” “Show me options near X, with Y, under Z.” Accor is betting the battle moves from ranking on Google → to being the best answer in AI chat. Why this is a big deal (even if bookings still happen on Accor.com) Because discovery is the choke point. If guests shortlist you inside AI: - you capture intent before OTAs and metasearch - you pull loyalty into the conversation earlier - you reduce friction to “book now” - you increase visibility where the next generation actually plans trips My hot take In 12–24 months, “we have an omnichannel inbox” will be table stakes. The differentiator will be: who runs the best AI-native guest journey: pre-stay personalization instant ops answers upsells at the right moment consistent brand tone zero friction handoffs to booking + staff

We talk a lot about "AI-driven experiences," but I wanted to show you what that actually looks like in reality. You shouldn't have to fill out a 10-field form just to invite a client to the office. Here is the workflow I just walked through using our AI, Kadence: 1. Natural Language is the new UI I opened Microsoft Teams (works in Slack too) and simply said: "I’ve got a guest coming in tomorrow at 3:00 PM... I need a meeting room for three people". No toggling apps. No calendar tetris. 2. Context is King The AI didn't ask me "Which building?" or "Which floor?" It already knows the floors I’m usually on and where my assigned space is. It immediately offered me relevant options on the 37th and 40th floors. 3. It Handles the "Busy Work" I added two external guests, "John and Jane." The system recognized their email domain (@taylor.com) wasn't part of my organization and automatically flagged them as visitors. Once I confirmed, it didn't just book the room. It automated the entire visitor journey: Sent Outlook/Google invites. Emailed guest parking and local transport instructions. Included the NDA and Wi-Fi details. 4. The "Contractor" Use Case I also tested booking a desk for a contractor named Frank. Because the system understands neighborhoods, it booked a desk near me automatically so we can actually collaborate. When they arrive? They scan a QR code at the kiosk, and I get alerted instantly. This is how simple it should be. I can’t promise the AI will bake you cookies (yet), but I can promise a delightful end-user experience

Do you want to build real-world machine learning skills and look like a data hero at work? I've got two words for you: Cluster Analysis Here's why it's your secret weapon: 1) Simple ML rules! Too many machine learning tutorials start with complex stuff—neural nets, deep learning, etc. But in the real world? Simple ML techniques that solve business problems fast are worth way more. Cluster analysis is one of those techniques. 2) But what is it? Clustering is an unsupervised machine technique that finds natural groupings in your data. No labels. No predictions. Just mining structure and insight directly from your data. 3) Why does this matter? Because most of the world's data is unlabeled and can't be used for predictive models. **AND** More data than ever is waiting to be mined to provide actionable insights to business stakeholders. Clustering helps you uncover patterns like: Customer segments Product affinities Behavioral groupings Regional differences 4) Real-world example: A marketing manager uses clustering to segment customers by shopping behavior. Instead of blasting the same promo to everyone, the marketing manager targets by cluster segments mined from the data: "Loyalists" "Deal hunters" "High-spenders" "Seasonal shoppers" 5) Another example: A product manager (PM) clusters users based on feature usage. The PM reveals hidden user types: Not based on arbitrary personas Not based on limited surveys But based on actual data from thousands of users. The net result? A data-driven product roadmap. All thanks to clustering. Greetings! 👋 My name is Dave Langer. I post daily about DIY topics like: SQL Excel Python Machine Learning Are you ready to learn Python for analytics and data science for free? Click "View my newsletter" above to get started.

🚀 Building an AI Agent for Hospitality Automation (Part 2) Update on the project — I’ve now built a working UI and connected it fully with the backend. What started as APIs is now turning into something you can actually use end-to-end. 💻 What’s done in this phase • Built the frontend using React + Tailwind CSS • Integrated it with the FastAPI + LangChain backend • All core functionalities are now working from the UI: ✔ Create booking ✔ Check booking ✔ Update booking ✔ Cancel booking You can simply interact through the interface, and the AI agent handles everything behind the scenes. from understanding the request to updating MongoDB. It finally feels like a real product instead of just a backend system. 💡 One thing I focused on here was keeping the interaction simple, the idea is that even non-technical users should be able to manage bookings without navigating complex dashboards. 🔜 What’s next Now moving towards making this more production-ready: • Adding authentication (multi-user access & security) 🔐 • Introducing voice-based booking so users can speak instead of type 🎙️ • Improving overall UX and flow • Gradually shaping this into a CRM-friendly system The goal is to build something that can actually fit into real hospitality workflows, not just a demo. I’ll keep sharing the journey as I keep building this out. If you’re working on AI agents, voice interfaces, or real-world AI products, would love to connect 🤝 #AI #AIAgents #React #TailwindCSS #FastAPI #LangChain #MongoDB #VoiceAI #BuildInPublic #HospitalityTech

This ChatGPT feature will change travel & hospitality forever. The day AI stops recommending and starts selling is here. Until recently, ChatGPT could tell you what to buy. Now, it can sell it to you directly. No browser tabs. No booking engines. Just: “I want this.” → “Buy.” → Done. It’s called Instant Checkout, built with Stripe . At launch it’s limited to simple products, but the direction is clear AI is becoming a commerce layer, not just a conversation. Why this changes everything for hotels For years, hotels have been optimising for search: keywords, metasearch, OTA rankings, paid clicks. That era is ending. Soon, travellers won’t “look” for hotels they’ll ask an assistant to design their stay. “Find me a design-led hotel in Florence with great coffee and an outdoor tub.” “Book a weekend where I can switch off completely, yoga, silence, forest.” And ChatGPT won’t just show links. It will build the itinerary, compare inventory, and complete the purchase inside the chat. That means the question is no longer “How do we rank higher on Google?” BUT “How do we become understandable and buyable to AI?” What tomorrow could look like Instead of booking engines, imagine this flow: Guest: “Plan me a 2-night recharge in Tuscany.” ChatGPT: “Would you like thermal baths, vineyard spa, or forest retreat?” Guest: “Vineyard.” ChatGPT: “I’ve found 3 properties. One includes cold plunges and biodynamic dining, €890. Confirm?” That’s not science fiction it’s the logical next step of Instant Checkout. The entire funnel collapses into a single dialogue. How hotels can prepare right now 1️⃣ Make your experiences machine-readable Structure your offers like data. “Private wine tasting, €120, 60 min, available Tue–Sat, includes transfer.” 2️⃣ Rethink your product catalogue Instead of “rooms” and “rates,” design modular experiences: morning rituals, energy resets, chef-led tastings, sunset rituals. 3️⃣ Expose your inventory to AI Ensure your website and PMS feed clear, structured information that APIs and agents can read and understand. 4️⃣ Keep ownership of fulfilment When bookings happen through assistants, whoever fulfils keeps the relationship don’t outsource that part. Hospitality used to be about visibility. Now it’s about readability. The hotels that speak the language of algorithms clear, structured, meaningful experiences will appear first when AI plans your guest’s next trip. AI won’t replace hoteliers. But it will reward the ones who make their experiences easy to buy. P.S. If you found this valuable, that’s exactly what we do. We help hotel brands refine their positioning, branding and user journey to drive direct bookings. Send me a DM for more info :)

One of my favorite parts of my job is working with customers on network and application architectures that operate at truly massive scale. And this one was especially fun! I had the chance to work with the team at Adobe on an architecture supporting Adobe Experience Platform running on Amazon Web Services (AWS). Adobe Experience Platform is the data foundation behind many modern digital experiences. It brings together massive volumes of customer data to power real-time insights, audience segmentation, and personalized journeys across channels. Building systems like this requires solving some interesting problems: • ingesting and processing high volumes of data • scaling infrastructure reliably for global workloads • enabling real-time insights and personalization • maintaining strong security and governance across the stack The case study walks through how Adobe built on AWS to deliver this at enterprise scale. Really proud to have been part of the architecture discussions behind this. If you're curious about what operating customer experience platforms at hyperscale actually looks like, this one is worth a read! https://lnkd.in/gasXZ6wc

4 practical ways to use AI in Product Marketing with prompts ⬇️ Most PMMs already use AI for content/idea generation and meeting summaries. But what are some other ways? There are a million applications, but for a brief LinkedIn post, here are some ideas and how-to steps: 1️⃣ Customer segmentation and messaging personalization How-to steps: ➖Collect customer information from CRM, website analytics, and sales data. ➖Establish criteria for segmentation (demographics, purchase history). ➖Use AI to sort customers into distinct groups and recommend tailored messaging. Prompts: ✅ “Analyze this customer dataset to identify segments based on behaviors like purchase history, engagement level, and product preferences.” ✅ “For each segment, create a profile with demographics, interests, and recommended messaging strategies.” 2️⃣ Sentiment analysis for products/features How-to steps: ➖Collect feedback from various sources, such as social media, product reviews, and customer service interactions. ➖Use AI to assess customer sentiment around products or features. ➖Highlight positives and improvement areas to inform messaging and product strategy. Prompts: ✅ “Analyze recent customer reviews of [Product Name] for sentiment around features like [Feature 1] and [Feature 2].” ✅“Identify top positive and negative themes in customer feedback, focusing on usability, performance, and support.” 3️⃣ Competitive intelligence How-To Steps: ➖Set up AI to monitor competitor websites, social media, and industry news. ➖Define key metrics for tracking. ➖Use AI to summarize competitive data, identifying trends. Prompts: ✅ “Analyze recent marketing campaigns from our top competitors. Identify common themes and unique selling points.” ✅“Compare our product features with those of [Competitor A] and [Competitor B]. Highlight areas where we excel or need improvement.” 4️⃣ Customer journey mapping How-To Steps: ➖Gather data on customer interactions across touchpoints (website, email, social media, support). ➖Use AI to identify common paths to sale and pain points in the customer journey. ➖Use recommendations to improve touchpoints and customer experience. Prompts: ✅ “Analyze our customer interaction data to identify the most common paths to purchase for [Product X].” ✅ “Based on customer behavior data, suggest improvements for our onboarding process to increase user activation.” To achieve this, you need more than ChatGPT or Claude. So, for tools to assist; I like Pendo for customer journey, Klue for competitor intelligence, and SalesForce Einstein for CRM segmentation help. 

At the National Retail Federation, we shared how Google is helping retailers move into the era of agentic commerce, with a clear focus on improving real customer experiences, not adding complexity.  And with that in mind, Sundar Pichai announced Gemini Enterprise for Customer Experience. Here’s what it’s all about: Today, shopping and customer service are fragmented. Customers repeat themselves across channels, and businesses manage disconnected systems. Gemini Enterprise for Customer Experience brings everything together into one intelligent, continuous journey. For customers, this means faster, more personalized interactions with less friction. For businesses, it means higher conversion, stronger loyalty, and better service at scale. Built specifically for retail, this platform allows AI agents to understand intent, keep context, and take action, from product discovery to checkout and post-purchase support, while retailers remain in control of the customer relationship. This is only one way we’re helping retailers turn every interaction into a long-term growth opportunity. If you want to read about what else Google and Google Cloud are doing to help businesses grow, read the full list of announcements here: https://bit.ly/4jEwbSZ. #GoogleCloud #NRF #GeminiEnterprise

𝐒𝐮𝐩𝐞𝐫𝐯𝐢𝐬𝐞𝐝 𝐂𝐥𝐮𝐬𝐭𝐞𝐫𝐢𝐧𝐠: 𝐰𝐡𝐞𝐧 𝐬𝐞𝐠𝐦𝐞𝐧𝐭𝐚𝐭𝐢𝐨𝐧 𝐬𝐭𝐚𝐫𝐭𝐬 𝐰𝐢𝐭𝐡 𝐦𝐨𝐝𝐞𝐥 𝐥𝐨𝐠𝐢𝐜, 𝐧𝐨𝐭 𝐫𝐚𝐰 𝐝𝐚𝐭𝐚 We usually think of SHAP values as a way to interpret how a model makes its predictions: for a single case or for a batch of them. But their use goes far beyond explainability. 💡 Imagine you have a customer segmentation problem: for churn, growth, or upsell analysis. You could apply a standard unsupervised clustering approach using selected features. But 𝐢𝐭 𝐝𝐨𝐞𝐬𝐧'𝐭 𝐫𝐞𝐟𝐥𝐞𝐜𝐭 𝐡𝐨𝐰 𝐢𝐦𝐩𝐨𝐫𝐭𝐚𝐧𝐭 𝐞𝐚𝐜𝐡 𝐟𝐞𝐚𝐭𝐮𝐫𝐞 𝐢𝐬 𝐟𝐨𝐫 𝐲𝐨𝐮𝐫 𝐛𝐮𝐬𝐢𝐧𝐞𝐬𝐬 𝐨𝐮𝐭𝐜𝐨𝐦𝐞, nor how features interact in the given problem's context, and especially not on an individual level. That's where SHAP values come to rescue. They let you look at each customer's state vector through the lens of the predictive model, grouping customers by similar prediction paths rather than by raw data values. This idea is known as 𝐒𝐮𝐩𝐞𝐫𝐯𝐢𝐬𝐞𝐝 𝐂𝐥𝐮𝐬𝐭𝐞𝐫𝐢𝐧𝐠, a concept introduced in the paper "𝐂𝐨𝐧𝐬𝐢𝐬𝐭𝐞𝐧𝐭 𝐈𝐧𝐝𝐢𝐯𝐢𝐝𝐮𝐚𝐥𝐢𝐳𝐞𝐝 𝐅𝐞𝐚𝐭𝐮𝐫𝐞 𝐀𝐭𝐭𝐫𝐢𝐛𝐮𝐭𝐢𝐨𝐧 𝐟𝐨𝐫 𝐓𝐫𝐞𝐞 𝐄𝐧𝐬𝐞𝐦𝐛𝐥𝐞𝐬" (Lundberg et al., 2018). Instead of clustering raw data, we 1️⃣ train a predictive model (e.g., gradient boosting), 2️⃣ then compute SHAP values: how much each feature contributed to the prediction for each observation. Then, we cluster these SHAP values, so each cluster 𝐠𝐫𝐨𝐮𝐩𝐬 𝐨𝐛𝐣𝐞𝐜𝐭𝐬 𝐭𝐡𝐚𝐭 𝐠𝐨𝐭 𝐬𝐢𝐦𝐢𝐥𝐚𝐫 𝐩𝐫𝐞𝐝𝐢𝐜𝐭𝐢𝐨𝐧𝐬 𝐟𝐨𝐫 𝐬𝐢𝐦𝐢𝐥𝐚𝐫 𝐫𝐞𝐚𝐬𝐨𝐧𝐬. 𝐖𝐡𝐲 𝐢𝐭'𝐬 𝐩𝐨𝐰𝐞𝐫𝐟𝐮𝐥: ✅ Same scale for all features as SHAP values are measured in the model's output units, solving the feature weighting problem. 🧭 Better interpretation as clusters reflect model logic, not just data similarity. 🧩 Actionable insights: you can identify subgroups (customers, patients, transactions) that the model treats in a similar way. 𝐓𝐡𝐢𝐬 𝐭𝐞𝐜𝐡𝐧𝐢𝐪𝐮𝐞 𝐰𝐨𝐫𝐤𝐬 𝐞𝐬𝐩𝐞𝐜𝐢𝐚𝐥𝐥𝐲 𝐰𝐞𝐥𝐥 𝐰𝐡𝐞𝐧: ▶️ You already have a strong predictive model. ▶️ You want to explore patterns behind the predictions. ▶️ You need interpretable clusters for strategy or communication. 👉 Supervised Clustering is a great way to connect prediction and segmentation, especially in business and healthcare use cases. #MachineLearning #DataScience #AdvancedML #ExplainableAI #SHAP #Clustering #PredictiveAnalysis