Understanding Technological Evolution

AI is rapidly moving from passive text generators to active decision-makers. To understand where things are headed, it’s important to trace the stages of this evolution. 1. 𝗟𝗟𝗠𝘀: 𝗧𝗵𝗲 𝗘𝗿𝗮 𝗼𝗳 𝗟𝗮𝗻𝗴𝘂𝗮𝗴𝗲 𝗙𝗹𝘂𝗲𝗻𝗰𝘆 Large Language Models (LLMs) like GPT-3 and GPT-4 excel at generating human-like text by predicting the next word in a sequence. They can produce coherent and contextually appropriate responses—but their capabilities end there. They don’t retain memory, they don’t take actions, and they don’t understand goals. They are reactive, not proactive. 2. 𝗥𝗔𝗚: 𝗧𝗵𝗲 𝗔𝗴𝗲 𝗼𝗳 𝗖𝗼𝗻𝘁𝗲𝘅𝘁-𝗔𝘄𝗮𝗿𝗲 𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝗼𝗻 Retrieval-Augmented Generation (RAG) brought a major upgrade by integrating LLMs with external knowledge sources like vector databases or document stores. Now the model could retrieve relevant context and generate more accurate and personalized responses based on that information. This stage introduced the idea of 𝗱𝘆𝗻𝗮𝗺𝗶𝗰 𝗸𝗻𝗼𝘄𝗹𝗲𝗱𝗴𝗲 𝗮𝗰𝗰𝗲𝘀𝘀, but still required orchestration. The system didn’t plan or act—it responded with more relevance. 3. 𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗔𝗜: 𝗧𝗼𝘄𝗮𝗿𝗱 𝗔𝘂𝘁𝗼𝗻𝗼𝗺𝗼𝘂𝘀 𝗜𝗻𝘁𝗲𝗹𝗹𝗶𝗴𝗲𝗻𝗰𝗲 Agentic AI is a fundamentally different paradigm. Here, systems are built to perceive, reason, and act toward goals—often without constant human prompting. An Agentic system includes: • 𝗠𝗲𝗺𝗼𝗿𝘆: to retain and recall information over time. • 𝗣𝗹𝗮𝗻𝗻𝗶𝗻𝗴: to decide what actions to take and in what order. • 𝗧𝗼𝗼𝗹 𝗨𝘀𝗲: to interact with APIs, databases, code, or software systems. • 𝗔𝘂𝘁𝗼𝗻𝗼𝗺𝘆: to loop through perception, decision, and action—iteratively improving performance.    Instead of a single model generating content, we now orchestrate 𝗺𝘂𝗹𝘁𝗶𝗽𝗹𝗲 𝗮𝗴𝗲𝗻𝘁𝘀, each responsible for specific tasks, coordinated by a central controller or planner. This is the architecture behind emerging use cases like autonomous coding assistants, intelligent workflow bots, and AI co-pilots that can operate entire systems. 𝗧𝗵𝗲 𝗦𝗵𝗶𝗳𝘁 𝗶𝗻 𝗧𝗵𝗶𝗻𝗸𝗶𝗻𝗴 We’re no longer designing prompts. We’re designing 𝗺𝗼𝗱𝘂𝗹𝗮𝗿, 𝗴𝗼𝗮𝗹-𝗱𝗿𝗶𝘃𝗲𝗻 𝘀𝘆𝘀𝘁𝗲𝗺𝘀 capable of interacting with the real world. This evolution—LLM → RAG → Agentic AI—marks the transition from 𝗹𝗮𝗻𝗴𝘂𝗮𝗴𝗲 𝘂𝗻𝗱𝗲𝗿𝘀𝘁𝗮𝗻𝗱𝗶𝗻𝗴 to 𝗴𝗼𝗮𝗹-𝗱𝗿𝗶𝘃𝗲𝗻 𝗶𝗻𝘁𝗲𝗹𝗹𝗶𝗴𝗲𝗻𝗰𝗲.

AI didn’t happen overnight, and it’s not one single concept. It’s the result of decades of progress - each breakthrough paving the way for the next. Here’s how the key building blocks fit together in the evolution of AI: 𝗔𝗿𝘁𝗶𝗳𝗶𝗰𝗶𝗮𝗹 𝗜𝗻𝘁𝗲𝗹𝗹𝗶𝗴𝗲𝗻𝗰𝗲 (𝗔𝗜) – technology that can analyse information, reason, and make context-based decisions without needing explicit instructions for every step. It’s the foundation for everything that followed. 𝗠𝗮𝗰𝗵𝗶𝗻𝗲 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 (𝗠𝗟)  – a branch of AI where systems learn from data instead of following fixed rules. They identify patterns and relationships in large datasets and adjust their behaviour accordingly. 𝗡𝗲𝘂𝗿𝗮𝗹 𝗡𝗲𝘁𝘄𝗼𝗿𝗸𝘀 (𝗡𝗡)  – a type of ML model inspired by the human brain. They’re especially good at recognising complex patterns, such as faces in photos, words in speech, or meaning in text. 𝗗𝗲𝗲𝗽 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 (𝗗𝗟)  – an advanced form of neural networks with many layers, trained on massive datasets. This made AI accurate enough for real-world use in language translation, image recognition, and voice assistants. 𝗣𝗿𝗲𝗱𝗶𝗰𝘁𝗶𝘃𝗲 𝗔𝗜  – the most common application of ML and DL today. It analyses historical data to predict what’s likely to happen next — from credit risk and demand forecasting to customer churn or fraud detection. 𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝘃𝗲 𝗔𝗜 (𝗚𝗲𝗻𝗔𝗜)  – a newer approach where AI doesn’t just analyse data but creates new content — writing text, generating images, coding, or composing music — based on what it has learned. 𝗔𝗜 𝗔𝗴𝗲𝗻𝘁𝘀  – autonomous applications that can make decisions and take actions on our behalf. They plan tasks, use other tools or systems, and complete goals with little or no human involvement. 𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗔𝗜 – a more advanced stage where multiple autonomous agents work together, share context, and make coordinated decisions to achieve broader goals. They don’t just execute tasks — they plan, adapt, and collaborate while remaining under human oversight. In reality, AI in its current form is really about extending human intelligence — and doing it at scale. Opinions: my own, Graphic sources: Gina Acosta Gutiérrez, Infinity Learning Subscribe to my newsletter: https://lnkd.in/dkqhnxdg

I just discovered why 90% of proptech sales fail, and it has nothing to do with the product's features. It's because founders don't understand how real estate developers actually make money. Let me show you the secret math that drives every decision they make. I was catching up with a proptech founder last week. His client, a GP, passed on software that would cost him $500/month. "They said it's too expensive!" he told me, frustrated. Then I showed him the math through the GP's eyes: $500/month = $6k/year = $120k hit to exit value (at a 5% cap rate) With his 20% promote, that's $24k straight out of his pocket. But here's where it gets interesting: Most vendors think real estate is about NOI. It's not. It's about the waterfall. Here's how it actually works: First, debt gets paid. Then, LPs get their principal back + preferred return (usually 8%). Only THEN does the GP get their promote (typically 20% of remaining profits). I used to tell founders: "Pitch the NOI increase!" Now I say: "Show them how to get past their pref faster." Different message. 10x the conversion. The promote is everything. It's why a GP will obsess over a $500/month expense but drop $50k on a lobby upgrade without blinking. One adds to NOI (and helps hit the promote). The other is just a cost. Want to sell into real estate? Stop thinking like a SaaS founder. Start thinking like a GP chasing a promote. Here's the framework I teach: • Calculate the NOI impact • Multiply by the exit cap rate • Show how it affects the promote • Watch them lean forward in their chair Example: "Your current vacancies cost you $10k/month in lost NOI. At a 5% cap, that's $2.4M in lost exit value. With your 20% promote, you're leaving $480k on the table." Now you're speaking their language. Most proptech founders think their enemy is the status quo. Wrong. Your enemy is the 8% pref. Every dollar matters. Every timeline matters. Every basis point matters. Because missing that promote doesn't just hurt the deal. It hurts the GP personally. I spent years watching smart operators pass on great solutions. Turns out they weren't cheap. They were doing math that the vendors didn't understand. Now I teach founders to lead with the waterfall. Sales cycles cut in half. The best prospects? Opportunistic developers 2 years from exit. The worst? Core owners collecting management fees. Different math. Different motivations. Different pitch. Stop selling software. Start selling promotes. P.S. If you want to master this (plus 50+ other frameworks for selling into real estate), we cover all of it in our course on 19th May. Join us- link in the comments. But honestly? This waterfall trick alone will transform your sales. Try it tomorrow. Thank me later.

At PwC, we've learned that the biggest barrier to scaling enterprise AI isn't model capability: it's trust. Here's how we think about that problem. Every new technology faces the same deadlock: you don't use it because you don't trust it, and you don't trust it because you don't use it. The way out is usually a trust proxy, a visible marker that tells people it's safe to change their behavior. The SSL padlock is the classic example. Ecommerce was technically possible in the 1990s, but adoption stalled because typing a credit card into a browser felt reckless. The padlock didn't create security, the encryption was already there. It made security visible. Enterprise AI faces the same issue. The models work. Real solutions exist. But capability is compounding faster than confidence. You see it in cautious adoption: professionals double-checking outputs the system got right. Not because the models aren't good enough, but because there's no structured way to show they've been rigorously evaluated by people who know what good looks like. These aren't capability problems. They're trust infrastructure problems. That's what we built Evaluation Navigator and the Human Alignment Center to address. 📊 Evaluation Navigator gives AI teams a consistent, repeatable way to evaluate solutions across the development lifecycle, with shared guidance and standardized reporting. By embedding evaluation directly into developer workflows through an SDK, trust markers are built into the solution as it's constructed, not stapled on before deployment. 🧐 The Human Alignment Center adds structured expert review at scale. Automated metrics can assess technical correctness, but in professional services the real question is whether the output reflects experienced professional judgment. The Human Alignment Center translates that judgment into dashboards and audit trails that governance leaders can actually act on. The padlock made invisible security visible. Evaluation infrastructure does the same for AI. Adoption is a trailing indicator of trust, so as evaluation becomes visible and accessible, adoption follows.

For years, one of the defining challenges in real estate was how slowly the industry adopted technology. In many ways, that lag is what created the opportunity for Fifth Wall in the first place: a massive, critical industry that sat out decades of software adoption and then had to start modernizing all at once. Even today, despite the growth of a real PropTech ecosystem, adoption is still slower and harder than in most other sectors. Historically, I saw that as a bug. A real constraint on innovation. What has changed is AI. Because so many real estate companies never fully embedded legacy enterprise software into their operations, they may now be in a better position to leapfrog directly into AI-native tools, workflows, and operating models. There is often less infrastructure to rip out, fewer entrenched systems to replace, and more room to build from scratch. That changes the equation. What used to look like resistance is starting to look more like flexibility. What used to look like a gap is starting to look more like a blank slate. And that is increasingly shaping how we think about the next wave of opportunity in real estate technology.

GenAI adoption is all about people, not about tools. Pharma giant Novo Nordisk offers a great case study of working out what supports useful uptake of AI across a large organization. A case study in MIT Sloan Management Review uncovers a range of useful lessons. Here are some of the most interesting. 🚀 Recognize a mid-cycle drop as normal. Novo Nordisk grew Copilot use from a few hundred to 20,000 users in just over a year, with 23% becoming frequent users within one month. However, by month three or four, 15% of early adopters dropped off and average time saved per week declined. Recognizing this dip as natural helped avoid panic and kept the focus on re-engagement strategies rather than getting staff to try tools for the first time. 🛠 Deliver function-specific training through champion networks. Generic AI onboarding failed to meet the needs of specialized roles. Novo Nordisk succeeded by creating domain-specific training, leveraging internal champions to contextualize AI use, and allowing teams to shape guidance based on their actual work. This addressed “AI shaming” and bridged confidence gaps across functions. 🤝 Use internal champions to overcome cultural resistance. Skepticism wasn’t solved by policy, it was shifted by influence. Novo Nordisk identified trusted, high-status employees to openly adopt and advocate for AI tools. Their visible endorsement encouraged hesitant peers to try AI without fear of judgment or failure. 📈 Treat adoption as a change process, not a tech rollout. Rather than pushing a one-time launch, Novo Nordisk framed GenAI as a long-term transformation. This meant investing in ongoing communication, support structures, and iterative learning. The approach acknowledged that adoption would ebb and flow, and prepared the organization to adapt accordingly. 🎯 Emphasize strategic value over time saved. Though average users saved about 2 hours per week, the most meaningful wins came from higher-quality work—more strategic thinking, clearer writing, and better planning. By highlighting these human-centric gains, Novo Nordisk built a stronger case for AI’s workplace relevance beyond mere productivity. 📊 Use employee data to shape the deployment strategy. Over 3,000 employee surveys and interviews helped Novo Nordisk spot where and why adoption lagged. This feedback guided real-time adjustments—like where to invest in new use cases, where to scale back, and how to tailor messaging. It also surfaced which functions became tool-reliant versus those needing more support.

New streaming data sources and AI’s use of them have revitalized the real-time event stream processing market and boosted revenue. Product leaders can use this research to assess how real-time data, analytics and AI can enhance and differentiate their offerings and adjust their roadmaps to leverage this potential. Gartner recommends that product leaders: 🔵 Allocate a portion of the engineering budget to evaluate the accessibility and applicability of real-time data and analytics that can impact desired business outcomes. Do so by experimenting with new data streams and event logs to understand their ability to inform and adapt products and services. 🔵 Work with engineering teams to design an architecture that can leverage real-time event stream data by identifying technology and requisite technology partnerships to consume the data within the reasonable confines of your product’s existing architecture. 🔵 Demonstrate the positive effect on decision quality and outcomes that result from including real-time contextual data in your products and services. Do so by measuring the accuracy of models that either predict outcomes or recommend actions, as well as embedding the best models in decision workflows. I asked Kevin R. Quinn, Vice President, Analyst - Technical Product Management, Gartner why he believe this research matters: 💡 "AI is accelerating every aspect of business. Decisions can’t just be based on what happened, but need to account for what is happening right now." 💡"Real-time data enables timely decision-making, enhances responsiveness, improves operational efficiency, and provides a competitive edge in rapidly changing environments." Our research shows how the market for real-time streaming data is changing, and how it is more accessible and relevant for providers and end-users, than ever before. Check out the insights from Kevin R. Quinn and myself (David Pidsley) which is exclusively available to Gartner clients who are product leaders subscribed to our "Emerging Technologies and Trends Impact on Products and Services" research. ▶️ "Emerging Tech: Revolutionize Your Products With Real-Time Data and AI" [Published 31 January 2025] 🔗 https://lnkd.in/ev7nk82R (requires client login) #DecisionIntelligence #RealTime #Data #AI #RealTimeData #StreamingData #StreamingAnalytics #StreamAnalytics #EventStream #EventStreamProcessing

Last week, a customer said something that stopped me in my tracks: “Our data is what makes us unique. If we share it with an AI model, it may play against us.” This customer recognizes the transformative power of AI. They understand that their data holds the key to unlocking that potential. But they also see risks alongside the opportunities—and those risks can’t be ignored. The truth is, technology is advancing faster than many businesses feel ready to adopt it. Bridging that gap between innovation and trust will be critical for unlocking AI’s full potential. So, how do we do that? It comes down understanding, acknowledging and addressing the barriers to AI adoption facing SMBs today: 1. Inflated expectations Companies are promised that AI will revolutionize their business. But when they adopt new AI tools, the reality falls short. Many use cases feel novel, not necessary. And that leads to low repeat usage and high skepticism. For scaling companies with limited resources and big ambitions, AI needs to deliver real value – not just hype. 2. Complex setups Many AI solutions are too complex, requiring armies of consultants to build and train custom tools. That might be ok if you’re a large enterprise. But for everyone else it’s a barrier to getting started, let alone driving adoption. SMBs need AI that works out of the box and integrates seamlessly into the flow of work – from the start. 3. Data privacy concerns Remember the quote I shared earlier? SMBs worry their proprietary data could be exposed and even used against them by competitors. Sharing data with AI tools feels too risky (especially tools that rely on third-party platforms). And that’s a barrier to usage. AI adoption starts with trust, and SMBs need absolute confidence that their data is secure – no exceptions. If 2024 was the year when SMBs saw AI’s potential from afar, 2025 will be the year when they unlock that potential for themselves. That starts by tackling barriers to AI adoption with products that provide immediate value, not inflated hype. Products that offer simplicity, not complexity (or consultants!). Products with security that’s rigorous, not risky. That’s what we’re building at HubSpot, and I’m excited to see what scaling companies do with the full potential of AI at their fingertips this year!

On why Service as Software startups will see more variability in revenue (less predictability) vs traditional SaaS (software as a service) startups. One of my portfolio companies in the rn red hot Service as Software (SAS) space raising a Series A, is asked a lot of questions about predictability of revenue (overall they have been on a tear but not all of their clients have expanded consistently). Now, I get why Series A (and some Series B investors) are keen on predictability. Predictability of revenue has been one of the canonical bedrock principles for evaluating quality of revenue in the SaaS world. Naturally, given that the SaaS model built around seat-based pricing and annual contracts lends itself well to predictability of revenue. But as we move from seat-based pricing (SaaS) to outcome-based pricing / selling labour units (SAS = service as software), investors will have to prioritise predictability a tad less. Why? Because you are moving away from seat-based pricing where you are one or two hops / quarters removed from business performance, and moving to outcome-based / selling work units, the SAS co is now integrated more deeply into your customer’s business performance. Hell, if you are a true SAS co, you are in effect the customer’s labour force, isn’t it? All of the unpredictability and challenges of the customer’s business model will get reflected in your business performance in real time and not 1-2-3 quarters away as in SaaS’s annual contracted seat-based approach. Series A / growth investors evaluating SAS models should therefore adapt / shift their mental framework away from SaaS and what has worked for SaaS when they come to SAS. I am not saying dont ask questions around revenue predictabilty and repeatability / retention. Pls do, but when you see some unpredictability or rather variability of rev, do not instinctively jump to say the revenue is of poor quality. Instead marry this data with other metrics to get a sense of how integrated you are into the customer’s workflow / revenue / supply chain. That is what gives you a sense of the strength of the SAS startup. When you are deeply embedded in your customer’s workflow, all of their business revenue variability gets reflected in your revenue stream as well. Variability of rev (not unpredictability) in SAS is a feature, not a bug.

AI adoption in enterprises rarely follows a straight line. You can build a capable agent that solves a real problem and still find no one using it. One extra click from the usual process can become an inhibitor. A new window, and your DAU/WAU/MAU can tank. Adoption isn’t just about rolling out a tool; it’s about reshaping ingrained habits. Teams grow so comfortable with existing workflows that AI tools can initially feel like a liability rather than a productivity enhancer. The journey moves through three stages: adoption, adaptation, and transformation. Strategy often starts with the end state (transformation), but execution must begin with the first step: adoption. Each stage requires building trust, lowering friction, and proving value in small, tangible increments. Without that, even the most well-designed AI solutions risk becoming "shelfware". AI isn’t a solo game. It’s a team sport. One weak link, one reluctant user, can cause the whole purpose to fall flat. Success depends not just on technology but on shared conviction. Real transformation happens when every click, every process, and every team member feels like AI isn’t an extra step but the obvious next one. #ExperienceFromTheField #WrittenByHuman