Hugo Raaijmakers

🧠 𝗔𝗻 𝗜𝗻𝘁𝗲𝗿𝗲𝘀𝘁𝗶𝗻𝗴 𝗥𝗲𝗮𝗱! This new research on Digital Twin-GPT explores forecasting patient health trajectories. Here are few key quotes I’ve pulled out from the paper: 𝗧𝗵𝗲 𝗢𝗽𝗽𝗼𝗿𝘁𝘂𝗻𝗶𝘁𝘆 “Generative artificial intelligence is revolutionizing digital twin development, enabling virtual patient representations that predict health trajectories, with large language models (LLMs) showcasing untapped clinical forecasting potential.” “We developed the Digital Twin—Generative Pretrained Transformer (DT-GPT), extending LLM-based forecasting solutions to clinical trajectory prediction.” “DT-GPT leverages electronic health records without requiring data imputation or normalization and overcomes real-world data challenges such as missingness, noise, and limited sample sizes.” 𝗧𝗵𝗲 𝗥𝗲𝘀𝘂𝗹𝘁𝘀 “Benchmarking on non-small cell lung cancer, intensive care unit, and Alzheimer’s disease datasets, DT-GPT outperformed state-of-the-art machine learning models, reducing the scaled mean absolute error by 3.4%, 1.3% and 1.8%, respectively.” “It maintained distributions and cross-correlations of clinical variables and demonstrated explainability through a human-interpretable interface.” “Additionally, DT-GPT’s ability to perform zero-shot forecasting highlights potential advantages of LLMs as clinical forecasting platforms, proposing a path towards digital twin applications in clinical trials, treatment selection, and adverse event mitigation.” 𝗧𝗵𝗲 𝗟𝗶𝗺𝗶𝘁𝗮𝘁𝗶𝗼𝗻𝘀 “However, performance declines for specific acute events—e.g., severe hemoglobin drops or high leukocyte counts—highlighting the challenge of forecasting low-prevalence, high-variance outcomes.” “Future improvements will require methods that enhance sensitivity to high-risk events, such as tailored loss functions, anomaly detection, and integration of unstructured clinical data.” “Another established shortcoming of LLM-based models is their tendency to hallucinate, as well as recreating the biases from the underlying data. In our case, the hallucination could be reflected in explainability results not necessarily providing true answers.” 𝗧𝗵𝗲 𝗕𝗼𝘁𝘁𝗼𝗺 𝗟𝗶𝗻𝗲 “[LLM shortcomings are] a critical aspect for the medical domain, and we believe that a human-in-the-loop setup will be required, together with advanced training of clinicians on the use of LLM outputs.” ______________________________ (Source in the comments.)

110

17 commentaren

The new Gartner Hype Cycle for AI is out, and it’s no surprise what’s landed in the trough of disillusionment… Generative AI. What felt like yesterday’s darling is now facing a reality check. Sky-high expectations around GenAI’s transformational capabilities, which for many companies, the actual business value has been underwhelming. Here’s why.… Without solid technical, data, and organizational foundations, guided by a focused enterprise-wide strategy, GenAI remains little more than an expensive content creation tool. This year’s Gartner report makes one thing clear... scaling AI isn’t about chasing the next AI model or breakthrough. It’s about building the right foundation first. ☑️ AI Governance and Risk Management: Covers Responsible AI and TRiSM, ensuring systems are ethical, transparent, secure, and compliant. It’s about building trust in AI, managing risks, and protecting sensitive data across the lifecycle. ☑️ AI-Ready Data: Structured, high-quality, context-rich data that AI systems can understand and use. This goes beyond “clean data”, we’re talking ontologies, knowledge graphs, etc. that enable understanding. “Most organizations lack the data, analytics and software foundations to move individual AI projects to production at scale.” – Gartner These aren’t nice-to-haves. They’re mandatory. Only then should organizations explore the technologies shaping the next wave: 🔷 AI Agents: Autonomous systems beyond simple chatbots. True autonomy remains a major hurdle for most organizations. 🔷 Multimodal AI: Systems that process text, image, audio, and video simultaneously, unlocking richer, contextual understanding. 🔷 TRiSM: Frameworks ensuring AI systems are secure, compliant, and trustworthy. Critical for enterprise adoption. These technologies are advancing rapidly, but they’re surrounded by hype (sound familiar?). The key is approaching them like an innovator...  start with specific, targeted use cases and a clear hypothesis, adjusting as you go. That’s how you turn speculative promise into practical value. So where should companies focus their energy today? Not on chasing trends, but on building the capacity to drive purposeful innovation at scale: 1️⃣ Enterprise-wide AI strategy: Align teams, tech, and priorities under a unified vision 2️⃣ Targeted strategic use cases: Focus on 2–3 high-impact processes where data is central and cross-functional collaboration is essential. 3️⃣ Supportive ecosystems: Build not just the tech stack, but the enablement layer, training, tooling, and community, to scale use cases horizontally. 4️⃣ Continuous innovation: Stay curious. Experiment with emerging trends and identify paths of least resistance to adoption. AI adoption wasn’t simple before ChatGPT, and its launch didn’t change that. The fundamentals still matter. The hype cycle just reminds us where to look. Gartner Report:  https://lnkd.in/g7vKc9Vr #AI #Gartner #HypeCycle #Innovation

1.255

182 commentaren

𝗔𝗜 𝗦𝗵𝗼𝗿𝘁𝗰𝘂𝘁𝘀 𝗪𝗼𝗻’𝘁 𝗕𝘂𝗶𝗹𝗱 𝗬𝗼𝘂𝗿 𝗩𝗜𝗣 𝗣𝗿𝗼𝗴𝗿𝗮𝗺 I discussed this topic during a recent catch-up call with Adar Ziv, after noticing a recurring pattern across several projects. Many operators are now attempting to build, restructure, or optimise their VIP programs using AI and large language models rather than relying on experienced professionals (internal or external). On paper, it looks efficient, cheaper, faster, and scalable. However, poor execution, rushed articulation, and ill-informed deployments are resulting in shortcuts backfiring. 𝗪𝗵𝗮𝘁’𝘀 𝗯𝗲𝗶𝗻𝗴 𝗼𝗯𝘀𝗲𝗿𝘃𝗲𝗱 👉 LLMs interpret theory but cannot build practice. They replicate frameworks and logic but lack the contextual experience and judgement to apply them effectively. 👉 They cannot read emotion, intent, or nuance. AI struggles to recognise the subtle behavioural signals skilled VIP specialists instinctively pick up through conversation, tone, and relationship. 👉 Output often looks polished, but fails in live environments. The strategy decks, workflows, and communications may appear sound, yet collapse under operational realities and player interaction. I have seen this firsthand. One current client approached me 18 months ago to build their VIP structure. After some back-and-forth, they decided to try doing it internally, using AI to lead the process. After months of trial, frustration, and wasted budget, they have returned asking for the same help we initially scoped a year and a half ago. While we are grateful for their trust, those 18 months were not used efficiently. Their decision ultimately cost time, money, and momentum. 𝗙𝘂𝗿𝘁𝗵𝗲𝗿 𝗰𝗼𝗻𝘀𝗶𝗱𝗲𝗿𝗮𝘁𝗶𝗼𝗻𝘀 👉 Any insights pulled from an LLM are also accessible to competitors. When using a public model, there is no such thing as a “proprietary” AI strategy. 👉 AI has advanced VIP operations significantly over the last three years, particularly through personalisation tools, early detection systems, and predictive analytics. While these tools are genuinely transformative, they rely on the operator having the 𝗽𝗿𝗼𝗽𝗲𝗿 𝗳𝗼𝘂𝗻𝗱𝗮𝘁𝗶𝗼𝗻 𝗮𝗻𝗱 𝗵𝘂𝗺𝗮𝗻 𝗼𝘃𝗲𝗿𝘀𝗶𝗴𝗵𝘁. 👉 Your technology stack supports your VIP program. It is not your VIP program. Above all, and I say this with respect: 𝗡𝗼 𝗔𝗜 𝘁𝗼𝗱𝗮𝘆 𝗰𝗮𝗻 𝗱𝗲𝘀𝗶𝗴𝗻, 𝗯𝘂𝗶𝗹𝗱, 𝗮𝗻𝗱 𝗺𝗮𝗻𝗮𝗴𝗲 𝗮 𝗿𝗼𝗯𝘂𝘀𝘁, 𝗺𝘂𝗹𝘁𝗶-𝗹𝗮𝘆𝗲𝗿𝗲𝗱 𝗩𝗜𝗣 𝗳𝗿𝗮𝗺𝗲𝘄𝗼𝗿𝗸. Not yet. 👉 Experts build structure, emotional understanding, and adaptability. 👉 AI builds theory, templates, and surface-level logic. Combining both can create value, but replacing people with prompts will never work in VIP. Not now, and very likely not in the future. If you are reworking your VIP program, building a new model, training your teams, or simply refining your strategy, 𝗱𝗼𝗻’𝘁 𝗹𝗲𝘁 𝘁𝗲𝗰𝗵𝗻𝗼𝗹𝗼𝗴𝘆 𝗼𝗿 𝗔𝗜 𝗹𝗲𝗮𝗱 𝘆𝗼𝘂. 100% use it, but use it as support, not as a shortcut. It will backfire.

43

11 commentaren

𝐍𝐨 𝐚𝐫𝐜𝐡𝐢𝐭𝐞𝐜𝐭, 𝐧𝐨 𝐜𝐢𝐭𝐲. Yet, every week, 𝐛𝐢𝐥𝐥𝐢𝐨𝐧-𝐝𝐨𝐥𝐥𝐚𝐫 𝐟𝐢𝐫𝐦𝐬 try to build AI systems with no architect at all. They call it “strategy.” What’s actually missing is 𝐚𝐫𝐜𝐡𝐢𝐭𝐞𝐜𝐭𝐮𝐫𝐞. Everyone talks about 𝐀𝐈 𝐚𝐦𝐛𝐢𝐭𝐢𝐨𝐧. Almost no one talks about 𝐀𝐈 𝐚𝐫𝐜𝐡𝐢𝐭𝐞𝐜𝐭𝐮𝐫𝐞. 👉 𝟗𝟓 % 𝐨𝐟 𝐀𝐈 𝐩𝐫𝐨𝐠𝐫𝐚𝐦𝐦𝐞𝐬 𝐬𝐭𝐚𝐥𝐥 𝑛𝑜𝑡 𝑏𝑒𝑐𝑎𝑢𝑠𝑒 𝑜𝑓 𝑚𝑜𝑑𝑒𝑙𝑠, 𝑏𝑢𝑡 𝑏𝑒𝑐𝑎𝑢𝑠𝑒 𝑜𝑓 𝑚𝑖𝑠𝑠𝑖𝑛𝑔 𝑎𝑟𝑐ℎ𝑖𝑡𝑒𝑐𝑡𝑢𝑟𝑒. The first five steps look great on paper - and that’s exactly where most organisations stop. 1️⃣ 𝐋𝐞𝐚𝐝 𝐟𝐫𝐨𝐦 𝐭𝐡𝐞 𝐜𝐞𝐧𝐭𝐫𝐞, 𝐚𝐜𝐭 𝐟𝐫𝐨𝐦 𝐭𝐡𝐞 𝐞𝐝𝐠𝐞. AI impact grows when the CEO owns ambition, but value emerges where business and tech meet. 2️⃣ 𝐃𝐞𝐬𝐢𝐠𝐧 𝐟𝐨𝐫 𝐫𝐡𝐲𝐭𝐡𝐦, 𝐧𝐨𝐭 𝐩𝐫𝐨𝐣𝐞𝐜𝐭𝐬. Weekly AI reviews beat quarterly innovation updates. Always. 3️⃣ 𝐁𝐮𝐢𝐥𝐝 𝐣𝐨𝐢𝐧𝐭 𝐨𝐰𝐧𝐞𝐫𝐬𝐡𝐢𝐩. Each function must own a measurable AI outcome - that’s when transformation becomes real. 4️⃣ 𝐌𝐚𝐤𝐞 𝐭𝐞𝐜𝐡𝐧𝐨𝐥𝐨𝐠𝐲 𝐜𝐨𝐦𝐩𝐨𝐬𝐚𝐛𝐥𝐞. Reusable components. Common data foundations. That’s how scale becomes sustainable. 5️⃣ 𝐈𝐧𝐯𝐞𝐬𝐭 𝐢𝐧 𝐩𝐞𝐨𝐩𝐥𝐞, 𝐧𝐨𝐭 𝐩𝐥𝐚𝐭𝐟𝐨𝐫𝐦𝐬. Upskilling outlasts tooling. Always has, always will. 𝐈 𝐥𝐞𝐚𝐫𝐧𝐞𝐝 𝐭𝐡𝐢𝐬 𝐭𝐡𝐞 𝐡𝐚𝐫𝐝 𝐰𝐚𝐲 - 𝐞𝐯𝐞𝐧 𝐰𝐨𝐫𝐥𝐝-𝐜𝐥𝐚𝐬𝐬 𝐟𝐢𝐫𝐦𝐬 𝐬𝐭𝐚𝐥𝐥 𝐫𝐢𝐠𝐡𝐭 𝐡𝐞𝐫𝐞. Not for lack of will, but for lack of 𝐚𝐫𝐜𝐡𝐢𝐭𝐞𝐜𝐭𝐮𝐫𝐞. 6️⃣ 𝐆𝐨𝐯𝐞𝐫𝐧𝐚𝐧𝐜𝐞 𝐚𝐬 𝐨𝐩𝐞𝐫𝐚𝐭𝐢𝐧𝐠 𝐬𝐲𝐬𝐭𝐞𝐦. Not committees - living guardrails. Trust, compliance, and agility moving in one rhythm. 7️⃣ 𝐕𝐚𝐥𝐮𝐞 𝐫𝐞𝐚𝐥𝐢𝐬𝐚𝐭𝐢𝐨𝐧 𝐚𝐬 𝐝𝐢𝐬𝐜𝐢𝐩𝐥𝐢𝐧𝐞. A dedicated 𝐕𝐚𝐥𝐮𝐞 𝐎𝐟𝐟𝐢𝐜𝐞 linking AI to measurable outcomes - tracking ROI in real time. 𝐃𝐢𝐬𝐜𝐢𝐩𝐥𝐢𝐧𝐞 𝐬𝐭𝐚𝐫𝐭𝐬 𝐭𝐡𝐞 𝐣𝐨𝐮𝐫𝐧𝐞𝐲. 𝐆𝐨𝐯𝐞𝐫𝐧𝐚𝐧𝐜𝐞 𝐚𝐧𝐝 𝐯𝐚𝐥𝐮𝐞 𝐚𝐫𝐜𝐡𝐢𝐭𝐞𝐜𝐭𝐮𝐫𝐞 𝐬𝐜𝐚𝐥𝐞 𝐢𝐭. AI doesn’t fail because of technology. It fails without an architect. 𝐖𝐡𝐞𝐫𝐞’𝐬 𝐲𝐨𝐮𝐫 𝐨𝐫𝐠𝐚𝐧𝐢𝐬𝐚𝐭𝐢𝐨𝐧 𝐭𝐨𝐝𝐚𝐲 - 𝐬𝐭𝐞𝐩 𝟓 𝐨𝐫 𝐬𝐭𝐞𝐩 𝟕? #Leadership #AI #Governance #ValueCreation #DigitalTransformation #AIFactory

1.042

43 commentaren

Throwback Friday: "Technology is easy. People are hard." That closing line from Peter Van Laer’s keynote at CEO Summit 2025, last Friday, stuck with me. AI is not the goal; it is a lever. The real test is how leaders and organizations choose to adopt it. Technology accelerates. Culture, ethics and collaboration decide whether we are truly future-proof. It was inspiring to hear how BDO Belgium is becoming an AI-enabled organization, step by step, with people at the center. At Akkodis Benelux we share that mindset: scale AI from pilots to production while building trust, adaptability and human connection. Three mindsets leaders need now: • Soulful thinking: anchor choices in purpose and impact. • Critical thinking: separate signal from noise and challenge assumptions. • Integrative thinking: connect disciplines to turn scattered insights into results. Thank you Jan De Schepper, for another excellent edition! So, how are you making AI work for people in your organization? #CEOSummit2025 #TechnologyAndHumanity #Leadership #AI #Innovation #PeopleFirst #Akkodis #Benelux

60

1 commentaar