<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Adam Keresztes on Medium]]></title>
        <description><![CDATA[Stories by Adam Keresztes on Medium]]></description>
        <link>https://medium.com/@adam.keresztes?source=rss-be48e241b83d------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*wIBP-m2L3OZFlG-R3PVRjQ.png</url>
            <title>Stories by Adam Keresztes on Medium</title>
            <link>https://medium.com/@adam.keresztes?source=rss-be48e241b83d------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 20:24:21 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@adam.keresztes/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[The term “Data” is dividing us]]></title>
            <link>https://medium.com/@adam.keresztes/the-term-data-is-dividing-us-abc5708aeb6f?source=rss-be48e241b83d------2</link>
            <guid isPermaLink="false">https://medium.com/p/abc5708aeb6f</guid>
            <category><![CDATA[semanticweb]]></category>
            <category><![CDATA[knowledge-management]]></category>
            <category><![CDATA[data-management]]></category>
            <category><![CDATA[semantics]]></category>
            <category><![CDATA[knowledge-graph]]></category>
            <dc:creator><![CDATA[Adam Keresztes]]></dc:creator>
            <pubDate>Sun, 28 Sep 2025 14:45:02 GMT</pubDate>
            <atom:updated>2025-09-28T16:32:43.824Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*oj3I0ik4Vj6ThcFCah_ykw.png" /></figure><p><strong>Background and perspective</strong></p><p>After a summer break spent recharging with family, I’ve been reflecting on my journey since moving from the field of user experience into semantics and knowledge representation. It’s been a significant shift — one that started with a difficult decision, but has proven to be the right move.</p><p>For most of my career, I focused on designing, creating digital experiences, creating mockups and prototypes that stakeholders loved and agreed were exactly what we needed. But time and time again, we couldn’t build them — because the data wasn’t there. Not that year, not the next year either. Eventually, I realized: if I truly wanted to bring those ideas to life, I had to move upstream and start working on the foundation — the structured knowledge itself, the data and how it’s represented.</p><p>After talking it over with my wife Mari, she simply said, “You seem to enjoy this work, so go for it.” That encouragement gave me the push I needed to take the leap. That was about five years ago, right when we began the IKEA Knowledge Graph journey — a move that has shaped much of my work since.</p><p>Since then, I’ve learned a new language, new tools, and new ways of thinking — while never fully taking off my UX glasses. And through this perspective, I’ve observed patterns not just in my current work, but also from past roles and in conversations with peers across other companies and the wider community. One pattern stands out: the growing divide between “business” and “data.”</p><p><strong>The divide between data and business</strong></p><p>After spending time in this space, I see the disconnect more clearly than ever. In many organizations, we treat “business” and “data” as separate worlds. And the gap is growing wider. Or at least its getting more apparent in recent times.</p><p>There’s been a strong push to become more data-driven — which is good — but often at the cost of replacing people who understand the business. Having talked to peers and old collegues form other companies, a majority of us observe the increate of hiring more data scientists, engineers, governance leads, and AI specialists, while letting go of people with long experience in whatever business the company is involved in, knowledge about products, materials, customers. These people carry the memory of the business — what’s worked, what hasn’t, and why. Instead of capturing that knowledge, we now try to rediscover it through data analysis and models.</p><p><strong>A real example of the problem</strong></p><p>At one point, I reached out to someone whose role involved developing products. Their response: “Hi Adam, thanks for the invite — but <strong>I don’t work with data</strong>.” So I asked, “So you don’t work with defining interior styles, materials, patterns and design expressions?” They replied, “Well yes, that’s exactly what I do — it’s my area of responsibility. But I don’t work with data.”</p><p>That moment has stuck with me. It shows how the divide we’ve created makes people feel like they have to choose a camp — either you’re in the business, or you’re in the data. But that’s a false split. Data should serve the business, not the other way around. And here I want to make a important distinction, the difference between operational data and analytical data. Of course analytical data is great to gain insights but we need to ensure that we have the right operational data in place from the start.</p><p>In this post I am only talking about operational data. Data needed to build digital products, logic, business rulesor customer experiences.</p><p><strong>Data trying to solve things we already know</strong></p><p>I do believe AI can solve a lot by looking at existing data — and in some cases, using that data to represent knowledge is exactly the right approach. Patterns in past behavior, aggregated observations, and statistical models can reveal things we might otherwise miss.</p><p>But in many cases, this approach can’t fully capture the why. It can tell us what tends to happen, but not the reasoning, context, or human judgment behind it.</p><p>In one example, interacted with a incredibly capable data science team working on the question: “Where should a coffee table go in a living room?” They analyzed thousands of images to find patterns and train a model to answer that.</p><p>But we already know the answer — or at least we know that it depends. Who lives there? What’s the purpose of the room? How is the space used? For example, if a home has a sofabed, the position of the coffee table might change completely depending on the state of the room — in a socialising state, it might be in front of the sofa for easy access, but in a sleeping state, it might be moved aside to allow the bed to unfold.</p><p>Instead of engaging with people who understand those variations, the team relies on image recognition to produce the statistically most common placement — without the why.</p><p>And I get it. I understand why a team would pursue this method — because you can start from day one, without having to track down the right person in the organization. But this is exactly where the problem lies. These teams are <strong>only working with the data</strong>, without the knowledge or perspectives from the business. And when they do reach out, they often hear the same line I mentioned earlier: <em>“I don’t work with the data.”</em></p><p><strong>Data quality without context</strong></p><p>This divide also shows up in data quality work. Often, the individuals and roles responsible for managing data don’t have enough business context to know if it’s complete, accurate, or even relevant. They are doing their best but they may not fully understand the value that knowledge generates, the implications if its quality is poor, or the structured knowledge itself.</p><p>Meanwhile, the people with that context — experts in safety, sustainability, customer needs, or product design — don’t think of themselves as working with data.</p><p>As a result, no one owns the full picture. Gaps go unnoticed, opportunities are missed, and sometimes bad data leads to bad decisions with real consequences for the business and its customers. In some cases, gaps in the operational data at source have resulted in workarounds that are far more resource intensive further down the line.</p><p><strong>Our language is part of the problem</strong></p><p>We make the divide worse through how we talk about data. Instead of saying “product features,” “usage requirements,” or “customer preferences,” we say “data domains,” “columns”, “data tables” and “data governance.” That language makes the work sound distant and technical — something separate from the business, rather than part of it.</p><p>I know it’s ironic that I’ve used the word “data” throughout this post, even though I believe it’s one of the things dividing us. We reach for it by habit — but maybe that’s the problem. What if instead, we talked about <strong>Shared</strong> <strong>Formal Knowledge </strong>or similar? Even this suggestion doesn’t feel fully right as I read it out loud but it gets closer — the structured foundation of our business understanding, formal enough for systems and humans, and solid enough to build value on?</p><p>That framing shifts the focus back to the <em>knowledge itself</em> — the shared truths and decisions that the business runs on. It’s the human expertise we’ve made explicit and reliable, so it can be scaled, connected, and used consistently across the whole company.</p><p><strong>The AI effect</strong></p><p>I’m not an AI expert, but I’ve seen enough to notice a pattern. With the rise of LLMs, RAG pipelines, embeddings, and multi-modal models, there’s a growing belief that we can solve almost anything with AI and the data we already have. The thinking goes: “We have the tools, we have the data — we can figure it out.”</p><p>And let me be clear — I’m all in on AI. Hell, I even used AI to help me write this post. AI is already part of our everyday lives: in our homes, in our work, in our businesses. It’s powerful, useful, and transformative. Its part of my everyday personal and work life.</p><p>But not everything needs to be predicted. Not everything should be inferred. Much of what we’re trying to extract from models is already known by someone in the business. We just haven’t built ways to capture it or make it reusable.</p><p>When that knowledge isn’t represented in a structured way, AI is left to “guess” — and that’s when hallucinations happen. If we want AI to deliver real value, we need to feed it with clear, consistent, and well-structured knowledge. That means taking what people know — the why behind decisions, the dependencies, the rules — and making it explicit, machine-readable, and connected.</p><p>Structured knowledge doesn’t just make AI more accurate; it makes it more explainable, more trustworthy, and less dependent on massive amounts of historical data to “figure things out.” Without that foundation, even the best models are like brilliant students working from incomplete notes — they might sound convincing, but they’ll still get critical things wrong. And when they do, accountability is blurred. Was it the model’s fault? The data? Or the people who deployed it? Structured knowledge helps make that chain of responsibility clear.</p><p><strong>Why this work matters to me</strong></p><p>This is why my move into semantics and knowledge representation has felt meaningful. It’s helped me see that we don’t need to choose between “data” and “business” — we need to bridge them as much as possible and create meaningful overlap.</p><p>But it’s bigger than that. It’s not just about connecting data with business processes or commercial goals. It’s about connecting knowledge with people. With humanity. With the things we find important — our culture, our creativity, our history, our values, and our shared sense of meaning.</p><p>Because at its core, structured knowledge is not just numbers in a system or entries in a table. It’s an expression of how we see and shape the world. If we strip away the human context, we reduce it to something cold, mechanical, and disconnected. When we keep the human element in, structured knowledge becomes a bridge — a way to carry culture, art, traditions, and ideas across time and between people.</p><p>That’s why I believe the future is in bringing these worlds together. Not by making “business people” into “data people,” or “data people” into “business people,” but by building systems that let knowledge flow freely between them — and beyond them, into all parts of life where meaning lives.</p><p><strong>What can we do about it?</strong></p><ul><li><strong>Reconnect the roles.</strong> Data teams should work closely with people who hold domain knowledge — from the start.</li><li><strong>Use everyday terms.</strong> Say “safety requirements” or “mattress firmness,” not just “data attributes.”</li><li><strong>Capture context.</strong> Don’t rely only on inference — represent what people already know.</li><li><strong>Involve business experts.</strong> Not to make them data people, but to structure the data in ways that reflect their knowledge.</li><li><strong>Treat knowledge as infrastructure.</strong> Data and business roles should work as one system, not in silos.</li></ul><p><strong>Bringing the two worlds together</strong></p><p>Let’s stop dividing people into “data” and “business.” That’s not how value is created. Let’s bring people who know closer to the systems we build — and make sure our structured knowledge reflects what we already understand. And let’s remember: it’s not just about better business. It’s about making structured knowledge a living reflection of human understanding, creativity, and purpose.<br> <br> So let me end with a question: <strong>if “data” divides us, what term could bring us together?</strong> What would resonate with both business people and data people — something better than “operational data,” something that truly reflects the structured knowledge at the heart of our work?</p><p>I’d also love to hear your experiences — examples of good practices, past successes, or ideas for how business and data can be united more effectively.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=abc5708aeb6f" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>