<?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 Balz Zürrer on Medium]]></title>
        <description><![CDATA[Stories by Balz Zürrer on Medium]]></description>
        <link>https://medium.com/@balz.zuerrer?source=rss-91cac6b5caa6------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*8W0-YxFCaR44VUkGqeFdGg.jpeg</url>
            <title>Stories by Balz Zürrer on Medium</title>
            <link>https://medium.com/@balz.zuerrer?source=rss-91cac6b5caa6------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 00:27:23 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@balz.zuerrer/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[Are We Automating Away Our Future Leaders?]]></title>
            <link>https://medium.com/@balz.zuerrer/are-we-automating-away-our-future-leaders-84b7e32a5242?source=rss-91cac6b5caa6------2</link>
            <guid isPermaLink="false">https://medium.com/p/84b7e32a5242</guid>
            <category><![CDATA[digital-transformation]]></category>
            <category><![CDATA[business-strategy]]></category>
            <category><![CDATA[leadership]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[ai-strategy]]></category>
            <dc:creator><![CDATA[Balz Zürrer]]></dc:creator>
            <pubDate>Fri, 20 Nov 2026 15:47:35 GMT</pubDate>
            <atom:updated>2026-11-20T15:47:35.296Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*u2XomE_EN_BtyiL-g09Zmw.png" /><figcaption>Created with AI</figcaption></figure><p><em>AI-first companies are improving productivity. But are they also eliminating the people who should become their future leaders?</em></p><p>Imagine a company that stops hiring junior employees because AI can handle most of their tasks.</p><p>For the first two years, everything looks fine.</p><p>Costs decrease. Productivity improves. Senior employees deliver more with fewer resources.</p><p>Five years later, the company faces a different problem. There is nobody ready to replace its experienced people.</p><blockquote>As a CEO, I believe this is one of the less discussed risks of AI transformation.</blockquote><p>We spend considerable time calculating how AI can reduce costs and improve efficiency. But we rarely ask what happens to the next generation of professionals when the tasks they traditionally learned from disappear.</p><p><strong>Who will develop the experience our companies need five or ten years from now?</strong></p><h4><strong>The Work We Automate Is Often the Work That Teaches</strong></h4><p>Think about how people become experts.</p><p>Junior consultants start by researching industries, preparing presentations, and supporting client projects.</p><p>Junior developers learn through coding, debugging, testing, and understanding systems built by more experienced colleagues.</p><p>Much of this work is becoming easier to automate.</p><p>Tools such as Microsoft Copilot already support research, documentation, and analysis. Agentic AI could take over increasingly complex workflows.</p><p>From an efficiency perspective, this is good news.</p><p>But these tasks serve another purpose. <strong>They teach people how to recognize problems, understand complexity, and develop judgment.</strong></p><p>A developer debugging a system learns how technology behaves when something goes wrong. A consultant preparing a presentation learns to distinguish important information from irrelevant details.</p><p>These experiences accumulate. Eventually, they become expertise.</p><p>If we automate the work without replacing the learning opportunities, we may discover that efficiency comes at an unexpected cost.</p><h4><strong>The Problem Will Not Appear Immediately</strong></h4><p>Imagine a technology consulting company with 100 employees.</p><p>As enterprise AI capabilities improve, management significantly reduces junior recruitment.</p><p>Senior consultants can now handle more analysis, documentation, and preparation with AI assistance.</p><p>Project margins improve. Clients receive results faster.</p><p>For the next few years, the decision appears successful.</p><p>But five years later, some senior experts have moved into leadership positions or left the company.</p><p>The organization needs people who can manage complex projects, challenge client assumptions, and make difficult decisions.</p><p>Normally, some of that talent would have developed internally.</p><p>But the pipeline has become smaller.</p><p>The company now needs to recruit experienced specialists externally, potentially at higher cost and with greater competition.</p><p>This is a hypothetical example, but it illustrates an important management challenge.</p><blockquote><strong>A decision can improve today’s financial performance while weakening tomorrow’s capabilities.</strong></blockquote><p>When evaluating the business impact of AI, leaders need to consider both.</p><h4><strong>Experience Is Not Something You Can Download</strong></h4><p>AI can help less experienced employees access knowledge, evaluate alternatives, and solve problems that previously required substantial expertise.</p><p>That is a significant opportunity.</p><p>But access to knowledge and professional judgment are different things.</p><p>Consider a consultant managing a difficult client situation.</p><p>AI might suggest several technically sound solutions. But the consultant still needs to understand why the client is resisting a recommendation, recognize internal tensions, and decide when to challenge rather than agree.</p><p>That judgment develops through experience.</p><p>The same applies to leadership. You cannot become a strong manager simply by reading excellent management advice. You need to make decisions, deal with consequences, and learn how people respond.</p><p>AI can accelerate learning, but we should not assume that better tools automatically produce experienced professionals.</p><p>For me, this is an important part of AI leadership: understanding which capabilities technology can strengthen and which still require deliberate human development.</p><h4><strong>Rethinking the First Five Years of a Career</strong></h4><p>I do not believe we should protect traditional junior roles simply to preserve them.</p><p>If AI can eliminate repetitive work, companies should take advantage of it.</p><p>But an AI-first operating model also needs a new approach to developing talent.</p><p>I see four priorities.</p><p><strong>1. Give junior employees responsibility earlier</strong></p><p>If AI handles much of the preparation, younger professionals should spend more time understanding decisions and their consequences. Instead of only preparing a client presentation, they should participate in discussions about what it needs to achieve.</p><p><strong>2. Make mentoring part of the operating model</strong></p><p>Knowledge transfer cannot depend entirely on individual initiative. Senior employees should have clear responsibility for developing younger colleagues. A specialist who develops three capable professionals may create more long-term value than one who only maximizes personal productivity.</p><p><strong>3. Use AI to accelerate learning</strong></p><p>There is a difference between asking AI to complete a task and using AI to understand how to complete it. A junior consultant using Microsoft Copilot can accept a generated summary or examine the sources, challenge the conclusions, and develop an independent recommendation. The same technology can either reduce learning or accelerate it. As Agentic AI takes on more complex activities, this distinction will become increasingly important.</p><p><strong>4. Measure more than productivity</strong></p><p>Most AI initiatives focus on time saved, costs reduced, and output increased. These metrics matter. But leaders should also ask:</p><blockquote>How many employees are ready to lead complex projects? Are younger professionals developing independent judgment? Is critical knowledge being transferred?</blockquote><p>A credible assessment of AI’s Business Impact should include the organization’s ability to sustain its performance over time.</p><h4><strong>A CEO Must Think Beyond the Next Efficiency Gain</strong></h4><p>I understand why companies are reconsidering their hiring plans.</p><p>Enterprise AI creates genuine opportunities to improve productivity. Some roles will change significantly, and others may disappear.</p><p>But treating junior employees only as a cost category is short-sighted.</p><p>Today’s junior consultant may become the person who leads a major client relationship or manages an important business unit in five years.</p><p>That future contribution is difficult to capture in a spreadsheet.</p><p>There is also a broader risk.</p><p>If many companies reduce entry-level hiring simultaneously, they may eventually compete for a smaller pool of experienced professionals.</p><p>What looks rational for one company today could create a talent shortage across an entire industry tomorrow.</p><p>This is why I believe talent development must be part of AI transformation, not an afterthought.</p><p>Before reducing junior hiring, I would ask management team one simple question:</p><blockquote><strong>If we hire significantly fewer junior employees today, who will be ready to lead our most important projects five years from now?</strong></blockquote><p>Every CEO should have a credible answer.</p><p>If you’d like to continue the conversation, feel free to connect with me on LinkedIn: <a href="https://www.linkedin.com/in/balzzuerrer/">linkedin.com/in/balzzuerrer</a>.</p><p>As Group CEO of Online Group, I help organizations turn AI and digital transformation into measurable business outcomes. Learn more at <a href="https://www.online.ch/en/">online.ch</a></p><p>For more perspectives on AI, technology and leadership, listen to <strong>«shift — Das Signal im Rauschen»</strong>, our podcast exploring the ideas and decisions shaping business transformation.</p><p>Available in Swiss German on <a href="https://www.youtube.com/playlist?list=PLH70kG1NyibE">YouTube </a>(with German subtitles) and <a href="https://open.spotify.com/show/033PfO15H2YOis98RUhIme">Spotify</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*syUYahxK6ABNOmm0wpCWUw.jpeg" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=84b7e32a5242" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI Speed vs Strategic Patience: Knowing When to Wait]]></title>
            <link>https://medium.com/@balz.zuerrer/ai-speed-vs-strategic-patience-knowing-when-to-wait-4e3e5f00a1d5?source=rss-91cac6b5caa6------2</link>
            <guid isPermaLink="false">https://medium.com/p/4e3e5f00a1d5</guid>
            <category><![CDATA[digital-transformation]]></category>
            <category><![CDATA[leadership]]></category>
            <category><![CDATA[business-strategy]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[ai-strategy]]></category>
            <dc:creator><![CDATA[Balz Zürrer]]></dc:creator>
            <pubDate>Fri, 06 Nov 2026 11:26:22 GMT</pubDate>
            <atom:updated>2026-11-06T11:26:22.642Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Sg36v81aGncuLIuK38u5EA.png" /><figcaption>Created with AI</figcaption></figure><p><em>Why AI leadership requires different speeds for different decisions.</em></p><p>Most companies believe they need to move faster with AI.</p><p>Some do. But others are already moving too fast in the wrong places.</p><p>They launch pilots, test new tools and encourage experimentation. Meanwhile, the infrastructure, governance and organizational decisions needed to scale AI receive far less attention.</p><p>The result is often impressive activity with limited business impact.</p><blockquote><strong><em>The challenge is knowing where speed creates value and where it creates problems that will take months to fix.</em></strong></blockquote><p>For leaders, this requires a different approach to AI transformation. Not every decision deserves the same urgency.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*krew4Qajm7hWMHs6V2UiSA.jpeg" /></figure><h3><strong>The problem with moving fast everywhere</strong></h3><p>AI makes experimentation unusually accessible.</p><p>A marketing team can test content generation in days. Customer support can prototype an AI assistant. Operations can explore automation without committing to a major transformation program.</p><p>This is useful. Companies need room to experiment.</p><p>But experimentation can also become a comfortable place to stay.</p><p>A successful pilot creates momentum. Another team wants to try something similar. Soon, the organization has a growing collection of AI initiatives, each with its own tools, assumptions and technical setup.</p><p>Leadership sees progress because there is plenty to report.</p><p>Yet when the discussion shifts to measurable ROI, production readiness or operational ownership, the picture becomes less convincing.</p><blockquote><strong><em>A company can be very active in AI without becoming meaningfully better at using it.</em></strong></blockquote><p>At some point, leaders have to decide which experiments deserve investment and which should stop.</p><p>That decision is often more valuable than launching the next pilot.</p><h3><strong>Four stages, four different speeds</strong></h3><p>I find it useful to distinguish between four stages of AI transformation.</p><p>Each requires a different management approach.</p><h4><strong>1. Exploration: Move fast</strong></h4><p>The purpose of exploration is to learn what works.</p><p>Teams should be able to test ideas quickly, with limited budgets and clear boundaries.</p><p>A useful experiment should answer a specific question. Can AI reduce the time required for a process? Can it improve service quality? Can it help employees make better decisions?</p><p>If an experiment takes months just to get started, something is probably wrong.</p><p>But speed needs an endpoint.</p><blockquote><strong><em>Every pilot should lead to a decision: scale it, improve it or stop it.</em></strong></blockquote><p>Otherwise, experimentation becomes an activity rather than a source of learning.</p><h4><strong>2. Capability building: Slow down where necessary</strong></h4><p>This is where many AI programs encounter difficulties.</p><p>A prototype that works for ten employees is not automatically a system that can support an entire organization.</p><p>Data access must be controlled. Systems need to integrate. Performance has to be monitored. Security and compliance requirements become operational responsibilities.</p><p>These are not particularly exciting topics.</p><p>They are also difficult to ignore once something goes wrong.</p><p>Consider a company that has developed several promising internal AI assistants. Each uses different data sources, authentication methods and external tools.</p><p>Individually, the pilots may work well.</p><p>Scaling them independently, however, creates unnecessary complexity and risk.</p><p>A better decision may be to pause additional deployments and establish shared technical and governance standards first.</p><p>That can feel slow when competitors are announcing new AI initiatives every week.</p><p>But rebuilding fragmented systems later is rarely faster.</p><blockquote><strong><em>Patience at this stage protects the ability to move quickly later.</em></strong></blockquote><h4><strong>3. Strategic bets: Think before committing</strong></h4><p>Some AI opportunities are relatively straightforward productivity improvements.</p><p>Others can change how a business operates.</p><p>An AI assistant that helps employees find information is one thing. A system that automatically makes customer decisions, changes pricing or handles significant parts of a service process is another.</p><p>The second category requires more than a technology assessment.</p><p>Leaders need to understand what changes for employees, customers and the organization itself.</p><p>Who remains accountable for decisions? Which roles will evolve? What happens when the system makes a mistake? Which capabilities should the company own rather than depend entirely on external providers?</p><p>These choices can shape the operating model for years.</p><p>I would be cautious about committing to a major AI transformation simply because a technology has become available.</p><blockquote><strong><em>Technical feasibility is not sufficient justification for strategic investment.</em></strong></blockquote><p>The expected business value, organizational consequences and risks must support the decision.</p><p>Strategic patience means taking enough time to understand those consequences before committing resources.</p><p>It does not mean waiting for certainty. That rarely exists.</p><h4><strong>4. Scaled execution: Accelerate again</strong></h4><p>Once an AI initiative has demonstrated value, the foundations are ready and leadership has made a clear commitment, the pace should change.</p><p>This is no longer the moment for endless discussion.</p><p>Processes need to be redesigned. Employees need training. Responsibilities must be assigned. Performance needs to be measured against business outcomes.</p><p>And importantly, the organization must be willing to change existing ways of working.</p><p>Deploying AI without changing the process around it often produces disappointing results.</p><p>Imagine an AI system that reduces the time required to process a customer request from fifteen minutes to five.</p><p>That sounds like a significant productivity gain.</p><p>But if the request still passes through the same approval steps, waits in the same queues and requires the same manual handovers, much of the potential value disappears.</p><p>The technology has accelerated one task. The operating model has barely changed.</p><p>This is why scaled execution requires management attention, not just technical deployment.</p><blockquote><strong><em>Once the direction is clear, slow organizational decisions can become a bigger obstacle than the technology itself.</em></strong></blockquote><h3><strong>The leadership decision that matters most</strong></h3><p>The difficult part is recognizing when an initiative should move from one stage to another.</p><p>A promising experiment may still be unsuitable for production.</p><p>A technically mature solution may lack a convincing business case.</p><p>And a strategically important initiative may be ready to scale while leadership continues asking for another pilot.</p><p>These situations require different decisions.</p><p>I would rather see a company scale two valuable AI use cases properly than celebrate twenty experiments that never become part of daily operations.</p><p>Of course, the right number depends on the organization. The principle is to distinguish learning activity from operational progress.</p><p>Leaders should regularly review their AI portfolio and ask:</p><ol><li><strong>What are we still trying to learn?</strong> Keep those experiments small and fast.</li><li><strong>What has demonstrated enough value to justify stronger foundations?</strong> Invest in the capabilities required to scale.</li><li><strong>Which initiatives could fundamentally change our operating model?</strong> Give those decisions proper strategic attention.</li><li><strong>What is already ready for deployment?</strong> Remove organizational obstacles and execute.</li></ol><p>The answers should influence funding, ownership and management priorities.</p><p>Without those consequences, the framework becomes another presentation.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CJclcYWbYKgdMTQYe0qwig.jpeg" /><figcaption>Michał Cisło (AI Engineer) and Michael Kammerer (CTO)</figcaption></figure><h3><strong>Patience also means knowing what not to automate</strong></h3><p>Not every AI decision should be automated at the same speed.</p><p>When decisions affect customers, employees or finances, accountability must be clear before deployment.</p><blockquote><strong><em>AI can support decisions, but responsibility remains with the organization.</em></strong></blockquote><p>Leaders need to know who owns the outcome, how errors are detected and when human intervention is required.</p><p>Without that clarity, moving faster only increases risk.</p><h3><strong>Finding the right pace</strong></h3><p>AI transformation requires different speeds.</p><p>Move fast when experimenting. Be deliberate when building capabilities and making strategic commitments. Accelerate once the direction is clear.</p><p><strong>The challenge is knowing when to change pace.</strong></p><p>At your next AI portfolio review, ask:</p><blockquote><strong>Which initiative are we pushing forward before it is ready, and which one are we holding back despite having enough evidence to act?</strong></blockquote><p>If you’d like to continue the conversation, feel free to connect with me on LinkedIn: <a href="https://www.linkedin.com/in/balzzuerrer/">linkedin.com/in/balzzuerrer</a>.</p><p>As Group CEO of Online Group, I help organizations turn AI and digital transformation into measurable business outcomes. Learn more at <a href="https://www.online.ch/en/">online.ch</a></p><p>For more perspectives on AI, technology and leadership, listen to <strong>«shift — Das Signal im Rauschen»</strong>, our podcast exploring the ideas and decisions shaping business transformation.</p><p>Available in Swiss German on <a href="https://www.youtube.com/playlist?list=PLH70kG1NyibE">YouTube </a>(with German subtitles) and <a href="https://open.spotify.com/show/033PfO15H2YOis98RUhIme">Spotify</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*lQJh-aJizkJz62CZs_vsOg.jpeg" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4e3e5f00a1d5" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI Pilots Are Easy. Scaling AI Is Where It Gets Hard.]]></title>
            <link>https://medium.com/@balz.zuerrer/ai-pilots-are-easy-scaling-ai-is-where-it-gets-hard-3c98678fb4ae?source=rss-91cac6b5caa6------2</link>
            <guid isPermaLink="false">https://medium.com/p/3c98678fb4ae</guid>
            <category><![CDATA[ai-leadership]]></category>
            <category><![CDATA[agentic-ai]]></category>
            <category><![CDATA[enterprise-ai]]></category>
            <category><![CDATA[ai-first]]></category>
            <category><![CDATA[ai-transformation]]></category>
            <dc:creator><![CDATA[Balz Zürrer]]></dc:creator>
            <pubDate>Fri, 23 Oct 2026 14:05:20 GMT</pubDate>
            <atom:updated>2026-10-23T14:05:20.945Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*V9oP5ARFzpHR0GoHsrkCCQ.png" /><figcaption>Created with AI</figcaption></figure><p><em>Lessons from Belimo’s approach to integrating AI into business processes and scaling operations.</em></p><p>Most companies have already experimented with AI. Some have deployed tools across their organizations. Others are testing agents and exploring automation.</p><p>But moving from successful experiments to AI that reliably supports business operations is a different challenge.</p><p>I recently discussed this with Sven Maushake, CIO of Belimo, on my podcast <a href="https://open.spotify.com/episode/0H8Lpe6LYQRic2obWcDRxj"><strong>«shift — Das Signal im Rauschen»</strong></a>. Belimo is an interesting case because the company experiences AI from two perspectives: as a supplier of components used in data center cooling infrastructure and as an enterprise looking to apply AI within its own operations.</p><p>Our conversation highlighted several practical lessons for leaders navigating AI Transformation.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Gl06uUlxuVU7mUa-3SsjIw.jpeg" /><figcaption>Sven Maushake</figcaption></figure><h4><strong>1. The next challenge is industrializing AI</strong></h4><p>Belimo has already explored generative AI through different use cases and proofs of concept. The focus is now shifting toward integrating AI into existing systems and business processes.</p><p>Sven described an approach built around three technological layers:</p><ul><li><strong>Employee productivity:</strong> More than 700 employees already use Microsoft Copilot, supported by training, communities, and knowledge sharing.</li><li><strong>Enterprise AI agents:</strong> Company-specific agents, supported by shared data and integration capabilities, can connect information across applications.</li><li><strong>Existing business systems:</strong> AI capabilities embedded in ERP and CRM platforms provide another opportunity to improve processes.</li></ul><p>What I find particularly relevant is how these layers can work together.</p><p>An AI agent that combines information from SAP and CRM, for example, could support activities that currently require employees to move between systems and manually connect information.</p><p>This is where Agentic AI becomes interesting from a business perspective. The opportunity extends beyond improving individual tasks to changing how entire processes operate.</p><h4><strong>2. AI-first thinking requires process ownership</strong></h4><p>One point from our discussion deserves particular attention.</p><p>At Belimo, responsibility for business processes remains with the business functions. IT contributes technical expertise, integration capabilities, and support for developing solutions.</p><p>I believe this distinction matters.</p><p>If AI initiatives remain primarily within IT, there is a risk that organizations optimize technology without sufficiently reconsidering the underlying processes.</p><p>Consider how a company handles customer requests for proposals.</p><p>Employees may need to review technical specifications, compare requirements, retrieve product information, and prepare responses.</p><p>Belimo is already piloting AI applications to support this type of work.</p><p>The potential Business Impact comes from examining which steps could be supported or eventually handled by specialized agents, rather than simply making each manual activity slightly faster.</p><p>That also changes the discussion about responsibilities.</p><p>When certain routine activities are automated, organizations need to reconsider how employees spend their time and where their expertise creates the greatest value.</p><h4><strong>3. Standardization becomes more important as AI scales</strong></h4><p>Belimo’s international growth provides another useful lesson.</p><p>Processes that developed independently in different markets may work well locally. But differences between them can become an obstacle when a company needs to scale operations globally.</p><blockquote>AI adds another dimension to this challenge.</blockquote><p>Agents depend on access to reliable information, consistent data structures, and integration between systems. Without these foundations, scaling successful pilots becomes difficult.</p><p>Sven was quite open about this. Looking back, he would have invested earlier and more heavily in data management and process standardization.</p><p>That is worth remembering when discussing enterprise AI.</p><p>Organizations often focus on the latest model capabilities. Yet the quality of their existing processes and data may have a greater influence on what they can realistically implement.</p><p>The challenge is finding the right balance between global standardization and the local flexibility customers still expect.</p><h4><strong>4. AI Leadership means knowing where to focus</strong></h4><p>Another aspect of Belimo’s approach caught my attention.</p><p>Rather than treating AI as a separate, centralized strategy, the company integrates it into functional strategies, including IT, supply chain, and innovation.</p><p>This places AI closer to the business decisions it is supposed to support.</p><p>It also raises expectations for leadership.</p><p>Executives do not need to understand every technical detail of generative AI. But they do need enough practical understanding to recognize opportunities, assess limitations, and decide where investment makes sense.</p><p>And those decisions are becoming harder.</p><p>Technology capabilities are developing quickly, while meaningful organizational change takes time. Companies cannot realistically turn every new AI capability into another transformation project.</p><p>Prioritization is therefore essential.</p><p>An interesting technology demonstration is not sufficient justification for investment. There needs to be a credible connection to an operational problem, a process improvement, or a business outcome.</p><h4><strong>From experimentation to operational value</strong></h4><p>My conversation with Sven reinforced something I increasingly consider important in AI Transformation.</p><p>The technology is advancing faster than many organizations can adapt their processes and ways of working.</p><p>Belimo’s experience shows why the next phase requires more than access to capable models. It involves connecting systems, improving data quality, standardizing processes where necessary, and helping employees adapt to changing responsibilities.</p><p>For leaders pursuing an AI-first operating model, I would start with a practical exercise:</p><p>Choose one business process where employees still spend considerable time collecting information, transferring data between systems, or handling repetitive steps.</p><p>Examine the process end to end. Identify what AI could realistically change, what data it would require, and who would remain accountable for the outcome.</p><p>That exercise may reveal more about your organization’s readiness for AI than another round of isolated experiments.</p><p>If you’d like to continue the conversation, feel free to connect with me on LinkedIn: <a href="https://www.linkedin.com/in/balzzuerrer/">linkedin.com/in/balzzuerrer</a>.</p><p>As Group CEO of Online Group, I help organizations turn AI and digital transformation into measurable business outcomes. Learn more at <a href="https://www.online.ch/en/">online.ch</a></p><p>Interested in the full conversation? This episode of <strong>«shift — Das Signal im Rauschen»</strong> is available in Swiss German on <a href="https://www.youtube.com/watch?v=VBxJV0RKA54">YouTube </a>(with German subtitles) and <a href="https://open.spotify.com/episode/0H8Lpe6LYQRic2obWcDRxj">Spotify</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*syUYahxK6ABNOmm0wpCWUw.jpeg" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3c98678fb4ae" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Hidden Cost of Being First in AI]]></title>
            <link>https://medium.com/@balz.zuerrer/the-hidden-cost-of-being-first-in-ai-febc0d552ba4?source=rss-91cac6b5caa6------2</link>
            <guid isPermaLink="false">https://medium.com/p/febc0d552ba4</guid>
            <category><![CDATA[ai-first]]></category>
            <category><![CDATA[ai-leadership]]></category>
            <category><![CDATA[enterprise-ai]]></category>
            <category><![CDATA[agentic-ai]]></category>
            <category><![CDATA[ai-transformation]]></category>
            <dc:creator><![CDATA[Balz Zürrer]]></dc:creator>
            <pubDate>Fri, 09 Oct 2026 13:58:52 GMT</pubDate>
            <atom:updated>2026-10-23T13:44:57.394Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OL1wFWUm2Xrn85AR6vFmSg.png" /><figcaption>Created with AI</figcaption></figure><p><em>What a conversation with SWICA’s CIO reveals about AI Transformation, technology decisions, and the value of knowing when to wait.</em></p><p>Companies face a difficult choice with AI. Move too early, and you risk investing in technology that is not ready. Wait too long, and you may miss opportunities that competitors are already using.</p><p>I believe the ability to make this distinction is becoming an important part of AI leadership.</p><p>In the fifth episode of our podcast <a href="https://open.spotify.com/episode/22IQJz29zPCpRIVdYLS3Db"><strong>«shift — Das Signal im Rauschen»</strong></a>, I spoke with Fabian Ringwald, CIO and Executive Board member at Swiss health insurer SWICA. His perspective offers a useful alternative to the pressure many organizations feel to adopt every new technology immediately.</p><p>SWICA follows what Fabian calls a smart-fast-follower strategy. The idea is straightforward: adopt technology when it is mature enough to create value, without waiting until the opportunity has passed.</p><p>There are several lessons here for leaders navigating AI Transformation.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*jCIafFBaYLRtb8wgU3CQjg.jpeg" /><figcaption>Fabian Ringwald</figcaption></figure><h4><strong>1. The Right Timing Matters More Than Being First</strong></h4><p>Not every company needs to develop its own AI solutions or become an early adopter.</p><p>For organizations that primarily use technology rather than produce it, being first can create unnecessary obligations.</p><p>An internally developed solution needs ongoing maintenance, investment, and improvement. Meanwhile, specialized technology providers continue developing their products.</p><p>Fabian made an interesting point during our conversation: sometimes being the third or fourth customer of a new solution is a better position than being the first.</p><p>You can benefit from the experience of others while still adopting the technology early enough to create meaningful Business Impact.</p><p>This requires judgment. Waiting should be a deliberate decision, not an excuse for avoiding change.</p><h4><strong>2. AI Experiments Need an Exit Decision</strong></h4><p>One of the most useful examples from our conversation was an early generative AI experiment at SWICA.</p><p>Even before ChatGPT became publicly available, SWICA worked with a technology partner on a solution to transcribe and summarize customer consultations.</p><p>The transcription worked surprisingly well.</p><p>The summarization was another matter. The system sometimes misinterpreted information or added content that had not been discussed.</p><p>For a process where the results would be shared directly with customers, that level of reliability was not acceptable.</p><p>The experiment did not continue as originally intended.</p><p>Fabian described the principle behind such decisions as “<em>Kill Your Darlings.</em>”</p><p>I think this deserves more attention in Enterprise AI initiatives.</p><p>Organizations are often good at approving experiments. They are less comfortable stopping them, especially after investing time, money, and internal credibility.</p><p>But continuing a project simply because it has already consumed resources is not a sound technology strategy.</p><p>An experiment should produce evidence. Sometimes that evidence tells you to stop.</p><h4><strong>3. AI Readiness Starts Long Before AI Adoption</strong></h4><p>Another important aspect of SWICA’s approach is its data architecture.</p><p>When Fabian joined the company, SWICA adopted a data mesh approach. Rather than making business units dependent on a central team for every data request, the goal was to provide reliable data products that teams could use themselves.</p><p>At the time, this was a long-term architectural decision.</p><p>Today, its relevance has increased with AI.</p><p>AI applications depend on data that is accessible, understandable, and reliable. Without those foundations, even powerful models have limited practical value.</p><p>This is particularly relevant for companies pursuing an AI-first operating model.</p><p>Introducing AI tools is one part of the transformation. Building the organizational and data foundations that allow people to use them effectively is another.</p><p>The second often needs to happen well before the first.</p><h4><strong>4. Resilience Is More Practical Than Complete Independence</strong></h4><p>Our conversation also touched on digital sovereignty and dependence on global technology providers.</p><p>Fabian challenged the idea that companies should aim for complete technological independence.</p><p>Dependencies have always existed. Cloud computing has changed their nature, but it has not created the concept of dependency itself.</p><p>His preferred focus is resilience.</p><p>For example, if an organization needs to change cloud providers, can it regain access to its data and continue using it within a reasonable timeframe?</p><p>That is a concrete question leaders can evaluate.</p><p>It also changes how technology decisions are made. Rather than treating every external dependency as a problem, organizations can assess the risks and determine which options they need to preserve.</p><p>For Enterprise AI, where external platforms and providers play a significant role, this is an important distinction.</p><h4><strong>5. AI Leadership Means Putting Decisions Where Knowledge Exists</strong></h4><p>Fabian also made a point about leadership that extends beyond technology.</p><p>SWICA has a clearly defined organizational hierarchy. But that does not mean every decision should move upward through the organization.</p><p>Good decisions should be made as close as possible to the people who have the relevant knowledge.</p><p>This principle connects naturally to the company’s data mesh approach. Both depend on making information available and giving people the responsibility to act on it.</p><blockquote>AI adds another dimension.</blockquote><p>Leaders need enough direct experience with AI to understand its possibilities and limitations. They cannot evaluate every decision solely through presentations or reports.</p><p>At the same time, they do not need to become the most technically knowledgeable people in the organization.</p><p>Their responsibility is to create conditions in which informed decisions can be made.</p><h4><strong>The Decision Worth Revisiting</strong></h4><p>What I took from my conversation with Fabian is that successful AI Transformation requires a more selective approach to innovation.</p><p>Some technologies deserve early investment. Others need more time to mature. And some experiments should be stopped, even when the original idea was promising.</p><p>The same judgment applies to data architecture, technology dependencies, and organizational decisions.</p><p>For leaders, a useful exercise is to review the AI initiatives currently underway and ask:</p><blockquote>Which projects are we continuing because the evidence supports them, and which are we continuing mainly because we have already invested in them?</blockquote><p>The answer may reveal more about an organization’s AI readiness than the number of pilots it has launched.</p><p>If you’d like to continue the conversation, feel free to connect with me on LinkedIn: <a href="https://www.linkedin.com/in/balzzuerrer/">linkedin.com/in/balzzuerrer</a>.</p><p>As Group CEO of Online Group, I help organizations turn AI and digital transformation into measurable business outcomes. Learn more at <a href="https://www.online.ch/en/">online.ch</a></p><p>Interested in the full conversation? The fifth episode of <strong>«shift — Das Signal im Rauschen»</strong> is available in Swiss German on <a href="https://www.youtube.com/watch?v=w2oDeJpst9E">YouTube </a>(with German subtitles) and <a href="https://open.spotify.com/episode/22IQJz29zPCpRIVdYLS3Db">Spotify</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KmqzVtF4ZajR7TL3WrNr_w.jpeg" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=febc0d552ba4" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[100 AI Use Cases Are Not an AI Strategy]]></title>
            <link>https://medium.com/@balz.zuerrer/100-ai-use-cases-are-not-an-ai-strategy-3051dd31fa2c?source=rss-91cac6b5caa6------2</link>
            <guid isPermaLink="false">https://medium.com/p/3051dd31fa2c</guid>
            <category><![CDATA[ai-transformation]]></category>
            <category><![CDATA[enterprise-ai]]></category>
            <category><![CDATA[ai-leadership]]></category>
            <category><![CDATA[agentic-ai]]></category>
            <category><![CDATA[ai-first]]></category>
            <dc:creator><![CDATA[Balz Zürrer]]></dc:creator>
            <pubDate>Fri, 25 Sep 2026 14:31:06 GMT</pubDate>
            <atom:updated>2026-09-25T14:31:06.218Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fzmypcQ1eL5ABOHP3dspgw.png" /><figcaption>Created with AI</figcaption></figure><p><em>Enterprise AI creates business impact when leaders start with the process, not the model.</em></p><p>There is no shortage of AI ideas. That is becoming part of the problem.</p><p>When Daniel Fiechter, CIO and Managing Director at Vitra, joined me for the fourth episode of <a href="https://open.spotify.com/episode/5CXcljHpverKTFPL9Qgc97"><strong>«shift — Das Signal im Rauschen»</strong></a>, we discussed what happened after Vitra created an internal AI task force.</p><blockquote>More than 100 use cases emerged.</blockquote><p>That sounds like progress. But a long list of use cases is not an AI strategy.</p><p>Different departments can end up solving similar problems with different partners and different technology stacks. Experimentation creates learning, but without focus it can also create complexity.</p><p>The challenge for AI leadership is therefore shifting.</p><p>Leaders do not need more ideas. They need a better way to decide which ideas deserve to become part of the operating model.</p><h4><strong>Start with the problem, not the model</strong></h4><p>One point from my conversation with Daniel stood out because it sounds obvious, yet is still easy to ignore:</p><p>Do not begin by asking which AI model or agent to deploy.</p><p>First understand the problem.</p><p>What outcome needs to improve? How does the process work today? Where is the actual constraint? What data and knowledge are required?</p><p>Only then does the technology question become useful.</p><p>Sometimes the answer will be AI. Sometimes traditional automation will solve the problem more effectively.</p><p>That distinction matters because an AI-first company should not become an AI-everywhere company.</p><p>AI-first, in my view, means being willing to rethink how work gets done with the capabilities now available. It does not mean forcing AI into every workflow.</p><h4><strong>The technology zoo is a management problem</strong></h4><p>There is a predictable pattern in enterprise AI.</p><p>A team discovers a problem. It finds a specialized tool. Another department encounters a similar problem and chooses another solution. Both experiments may be perfectly reasonable in isolation.</p><p>Repeat this often enough and the organization accumulates a technology zoo.</p><p>More applications. More integrations. More cost. More knowledge distributed across separate systems.</p><p>And not necessarily more business impact.</p><p>Vitra is taking a platform-based approach to avoid this fragmentation. Its AI initiatives build on the Microsoft environment already used for documents and knowledge, including Microsoft 365, SharePoint and Teams.</p><blockquote>This changes the discussion.</blockquote><p>Instead of choosing a new technology stack for every use case, more attention can go toward the business problem and the process behind it.</p><p>That becomes increasingly important as Microsoft Copilot and other AI capabilities become part of everyday work.</p><h4><strong>AI Transformation requires process redesign</strong></h4><p>There is another trap I see in the way organizations think about AI.</p><p>They take an existing process, add AI somewhere in the middle, and call it transformation.</p><p>The process may become faster. That can be useful.</p><p>But it leaves a larger opportunity unexplored.</p><p>Daniel’s approach is to ask how the process itself could be redesigned using today’s technological possibilities.</p><p>That is a much harder question.</p><p>It forces leaders to examine assumptions that were built into workflows years ago.</p><p>Consider a process involving internal product knowledge.</p><p>The narrow question is:</p><blockquote><strong>Where can AI make the existing steps faster?</strong></blockquote><p>The broader question is:</p><blockquote><strong>If employees can access and work with knowledge differently, which of those steps are still necessary?</strong></blockquote><p>The second question can lead to a fundamentally different process.</p><p>That is where AI transformation starts becoming an operating-model discussion rather than a tooling discussion.</p><h4><strong>A practical framework for moving from ideas to impact</strong></h4><p>Based on our conversation, I would reduce the challenge to five decisions.</p><h4><strong>1. Define the business problem</strong></h4><p>Do not start with a model, an agent, or a feature.</p><p>Define the outcome you want to change.</p><p>At Vitra, the current focus includes areas such as efficiency, sales support and product knowledge management.</p><p>That creates boundaries. And boundaries are useful when new AI possibilities appear almost every day.</p><h4><strong>2. Understand the existing process</strong></h4><p>Before automating something, understand what actually happens.</p><p>Which steps create value? Which exist because of old technical limitations? Which require judgment?</p><p>Without that understanding, AI can simply automate unnecessary complexity.</p><h4><strong>3. Choose the simplest technology that works</strong></h4><p>Not every problem requires AI.</p><p>Traditional automation remains the right answer for some well-defined processes.</p><p>Enterprise AI should expand the set of available options, not remove the discipline of choosing between them.</p><h4><strong>4. Build on a shared foundation</strong></h4><p>Every isolated solution creates future decisions around integration, governance, knowledge and cost.</p><p>A platform approach can reduce some of that fragmentation.</p><p>For Vitra, the existing Microsoft environment provides that foundation. The objective is not to debate models repeatedly, but to make knowledge more accessible and build solutions on common infrastructure.</p><h4><strong>5. Decide where humans remain in control</strong></h4><p>This becomes particularly important with Agentic AI.</p><p>Daniel described a distinction leaders will increasingly have to make: which tasks can run autonomously, and which still require human decision-making?</p><p>Processes with clear outcomes can increasingly be automated.</p><p>Complex decisions, critical situations and work requiring context, experience or tact still require people.</p><p>That boundary needs to be designed. It should not emerge accidentally from technology adoption.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5tUxlwzGEpYbDoBfxTDnxg.jpeg" /><figcaption>Daniel Fiechter</figcaption></figure><h4><strong>Adoption needs both directions</strong></h4><p>AI transformation cannot be run only from the top.</p><p>But it cannot remain an experiment at the edges of the organization either.</p><p>Vitra currently has around 40 to 50 AI enthusiasts across different departments who test possibilities and share what they learn. Accessible training and regular exchange help spread practical experience, including the use of Copilot Chat for text, image, video and other AI applications.</p><p>That bottom-up energy matters.</p><p>It gives people room to discover where the technology is useful in actual work.</p><p>But when AI begins changing a core process, leadership has to make decisions.</p><p>Processes cross teams. They affect responsibilities. They change how work is organized.</p><p>At that point, experimentation needs direction.</p><p>This is one of the less visible parts of AI leadership:</p><blockquote>knowing when exploration should continue and when a decision needs to be made.</blockquote><h4><strong>Agentic AI may change what leadership measures</strong></h4><p>AI agents introduce another interesting question.</p><p>If part of a team’s work can be performed autonomously, how should we think about management?</p><p>Daniel made an important observation during our conversation: leadership may become less about how many employees someone manages and more about the results that are achieved, regardless of whether individual tasks are performed by humans or agents.</p><p>I think this is worth watching closely.</p><p>It shifts attention from organizational size toward outcomes.</p><p>It also puts more responsibility on leaders to define where autonomy is appropriate and where human involvement remains necessary.</p><p>The org chart will not answer that question.</p><p>The process will.</p><h4><strong>The next AI decision should probably be a process decision</strong></h4><p>The pace of AI development makes it tempting to keep adding.</p><p>Another model. Another use case. Another agent. Another tool.</p><p>But accumulating technology is not the same as building capability.</p><p>A useful test for the next AI initiative is simple:</p><blockquote>Before approving another AI tool, take one important process and redesign it from scratch using the capabilities you already have.</blockquote><p>Then ask what should be automated, what still requires human judgment, and what should disappear from the process entirely.</p><p>The answer may tell you more about your AI strategy than another 100 use cases ever could.</p><p>If you’d like to continue the conversation, feel free to connect with me on LinkedIn: <a href="https://www.linkedin.com/in/balzzuerrer/">linkedin.com/in/balzzuerrer</a>.</p><p>As Group CEO of Online Group, I help organizations turn AI and digital transformation into measurable business outcomes. Learn more at <a href="https://www.online.ch/en/">online.ch</a></p><p>Interested in the full conversation? The fourth episode of <strong>«shift — Das Signal im Rauschen»</strong> is available in Swiss German on <a href="https://www.youtube.com/watch?v=Zn5ZtdczbBo">YouTube </a>(with German subtitles) and <a href="https://open.spotify.com/episode/5CXcljHpverKTFPL9Qgc97">Spotify</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*syUYahxK6ABNOmm0wpCWUw.jpeg" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3051dd31fa2c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI-First Does Not Mean AI Everywhere]]></title>
            <link>https://medium.com/@balz.zuerrer/ai-first-does-not-mean-ai-everywhere-d2b052d37fcc?source=rss-91cac6b5caa6------2</link>
            <guid isPermaLink="false">https://medium.com/p/d2b052d37fcc</guid>
            <category><![CDATA[enterprise-ai]]></category>
            <category><![CDATA[agentic-ai]]></category>
            <category><![CDATA[ai-leadership]]></category>
            <category><![CDATA[ai-transformation]]></category>
            <category><![CDATA[ai-first]]></category>
            <dc:creator><![CDATA[Balz Zürrer]]></dc:creator>
            <pubDate>Fri, 11 Sep 2026 14:18:15 GMT</pubDate>
            <atom:updated>2026-09-11T14:18:15.141Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*k3p1hCv2gzpS-AvPkgqysw.png" /><figcaption>Created with AI</figcaption></figure><p><em>What a conversation about digitizing Liechtenstein taught me about focus, sovereignty and AI Leadership.</em></p><p>There is an uncomfortable side effect of the current “AI boom”.</p><p>It has become very easy to start something.</p><p>Another use case. Another pilot. Another tool. Another license.</p><p>The harder part is deciding what deserves to survive.</p><p>I recently spoke with Fabian Schmid, Director of the Liechtenstein Office of Information Technology in our <a href="https://open.spotify.com/episode/5H3vGbYvDz8Ait6eEvGZfP"><strong>«shift — The Signal in the Noise»</strong></a> podcast, about the digital transformation of an entire country. Our conversation covered AI, changing roles, cloud infrastructure, open source and digital sovereignty.</p><p>But one idea stayed with me.</p><blockquote>Liechtenstein does not try to follow every development in AI.</blockquote><p>Fabian calls their approach being a “smart follower.”</p><p>I think many organizations could benefit from the same discipline.</p><h4><strong>Being AI-first is not a race to adopt everything</strong></h4><p>New AI models and applications appear faster than most organizations can evaluate them properly.</p><p>Trying to keep up with all of them is not a strategy.</p><p>It is especially unrealistic for an organization with limited resources and a wide range of responsibilities. A government IT department does not serve one product or one business process. It supports different agencies, requirements and specialized applications.</p><p>So the approach in Liechtenstein is selective.</p><ul><li>Observe what works.</li><li>Choose what is relevant.</li><li>Implement it.</li><li>Then evaluate whether it actually helps.</li></ul><p>This sounds less exciting than announcing dozens of AI use cases. But it is much closer to how meaningful AI Transformation happens.</p><p>The number of ideas matters very little if they remain ideas.</p><p>What matters is whether a process becomes more efficient, a task is better supported, or a new service becomes possible.</p><p>That is business impact and it is a much harder standard than experimentation.</p><h4><strong>The license is where the work starts</strong></h4><p>There is another trap in Enterprise AI.</p><p>We confuse availability with adoption.</p><p>Give people access to an AI tool and technically, the rollout is complete. Operationally, very little may have changed.</p><p>Fabian made this point clearly. Employees need support, training and concrete examples. They need to understand how the technology applies to their actual work.</p><p>Microsoft Copilot is a good example of the broader challenge. Providing access to a tool does not automatically redesign a workflow.</p><p>Someone still has to ask what should now be done differently.</p><blockquote>Which tasks should change?</blockquote><blockquote>Where does AI genuinely help?</blockquote><blockquote>Where is human judgment still required?</blockquote><blockquote>And what should people do with the capacity that AI potentially creates?</blockquote><p>These are management questions, not software questions.</p><p>This is where AI leadership becomes visible. Not in how quickly a tool is purchased, but in whether leaders are willing to rethink the work around it.</p><h4><strong>Roles will move, even when job titles do not</strong></h4><p>Software development gives us an early indication of what this can look like.</p><p>Fabian expects developers to write less code themselves over time. Their role can shift toward orchestration, quality assurance and technical translation.</p><p>That does not make the developer irrelevant.</p><blockquote>It changes what being a good developer requires.</blockquote><p>Requirements become important. Architecture remains important. Security remains important. Interpretation remains important.</p><p>This becomes even more relevant as agentic AI expands what can be delegated to technology.</p><p>The more an AI system can do, the more precisely an organization needs to define what people remain responsible for.</p><p>I think this is where some AI Transformation programs will become uncomfortable.</p><p>Saving time is easy to celebrate.</p><p>Redesigning responsibility is harder.</p><p>If AI removes 20 percent of a person’s existing tasks, the interesting question is not only what happened to those tasks.</p><p>It is what should happen to the person’s role.</p><p>Organizations that avoid that discussion may automate parts of work without changing how the organization actually operates.</p><h4><strong>Sovereignty gets harder when you leave the PowerPoint</strong></h4><p>The same pragmatism appears in Liechtenstein’s approach to digital sovereignty.</p><p>There is no simple cloud versus on-premises answer.</p><p>The government uses Microsoft 365 for collaboration. Confidential and classified information, as well as key line-of-business applications, remain hosted in its own data centers.</p><p>Different information receives different treatment depending on the protection it requires.</p><p>That is a more useful way to think about sovereignty than searching for one technology choice that solves the entire problem.</p><p>The same applies to open source.</p><p>Moving to open-source technology does not automatically make an organization cheaper or more secure. Licensing is only one part of the equation.</p><p>Integration has to work. Systems have to be maintained. Security has to be managed. People need support and training. The operational model still has to function.</p><p>Digital sovereignty therefore has a price. Sometimes that price is worth paying. Sometimes a different balance makes more sense.</p><p>The important part is knowing which trade-off you are making.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*79KHMpYnGYkO1roBwcytow.jpeg" /><figcaption>Fabian Schmid</figcaption></figure><h4><strong>Perhaps AI-first should mean being more selective</strong></h4><p>An AI-first organization is not defined by how much AI it uses, but by how deliberately it uses it.</p><p>It knows which experiments deserve to become real capabilities, where human accountability remains, and which risks are worth taking.</p><p>A simple test:</p><blockquote>Pick one AI use case already in operation. Can you explain what changed in the actual work, who is accountable for the outcome, and what Business Impact it creates?</blockquote><p>If not, I would not add another use case yet.</p><p>If you’d like to continue the conversation, feel free to connect with me on LinkedIn: <a href="https://www.linkedin.com/in/balzzuerrer/">linkedin.com/in/balzzuerrer</a>.</p><p>As Group CEO of Online Group, I help organizations turn AI and digital transformation into measurable business outcomes. Learn more at <a href="https://www.online.ch/en/">online.ch</a></p><p>Interested in the full conversation? The third episode of «<strong>shift — The Signal in the Noise</strong>» is available in Swiss German on <a href="https://www.youtube.com/watch?v=Jh17NHQiiG8&amp;t=1s">YouTube </a>(with German subtitles) and <a href="https://open.spotify.com/episode/5H3vGbYvDz8Ait6eEvGZfP">Spotify</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*syUYahxK6ABNOmm0wpCWUw.jpeg" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d2b052d37fcc" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Your AI Strategy Has a People Problem]]></title>
            <link>https://medium.com/@balz.zuerrer/your-ai-strategy-has-a-people-problem-5591f6a58aa7?source=rss-91cac6b5caa6------2</link>
            <guid isPermaLink="false">https://medium.com/p/5591f6a58aa7</guid>
            <category><![CDATA[ai-leadership]]></category>
            <category><![CDATA[agentic-ai]]></category>
            <category><![CDATA[ai-first]]></category>
            <category><![CDATA[enterprise-ai]]></category>
            <category><![CDATA[ai-transformation]]></category>
            <dc:creator><![CDATA[Balz Zürrer]]></dc:creator>
            <pubDate>Fri, 28 Aug 2026 06:31:01 GMT</pubDate>
            <atom:updated>2026-08-28T06:31:01.678Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ogKgReGI9AXHwp1W1Q3U2Q.png" /><figcaption>Created with AI</figcaption></figure><p>Most companies have understood that AI matters.</p><p>That is no longer the difficult part.</p><p>The difficult part is moving from individual tools and pilots to a company that actually works differently because of AI.</p><p>In the second episode of our <a href="https://open.spotify.com/episode/02Np1olGrbJ7trcIRpsnJc">«<strong>shift — The Signal in the Noise</strong>»</a> podcast, I discussed this with Marco Denzler, CEO Switzerland at Online Group. One point from that conversation stayed with me:</p><blockquote>Becoming AI-first is as much a people question as a technology question.</blockquote><p>And that has consequences for how leaders think about transformation, skills, and even recruiting.</p><h3><strong>AI-first is not a tool rollout</strong></h3><p>It is tempting to define AI adoption by what has been deployed.</p><p>A new assistant. A pilot. An AI feature inside an existing application.</p><blockquote>These are useful steps, but they do not make a company AI-first.</blockquote><p>For us, AI-first means that people actively explore where AI can improve their daily work. That applies across development, consulting, back-office operations, and marketing.</p><p>The important distinction is between having access to AI and changing how work gets done with AI.</p><p>That is where AI Transformation starts to become an operating question.</p><h3><strong>A simple test for AI Transformation</strong></h3><p>I would look at four areas.</p><p><strong><em>1. Are people actually using AI in their work?</em></strong></p><p>Providing technology is the easy part.</p><p>People need an environment where curiosity, learning, and experimentation are encouraged. They need to be able to test what works, understand what does not, and improve how they use AI over time.</p><p><strong>Important</strong>: One training session will not solve this.</p><p>The technology is developing too quickly for learning to be treated as a one-off event.</p><p><strong><em>2. Is AI connected to business processes?</em></strong></p><p>Many companies are still experimenting.</p><p>They have initial tools and pilots, but the connection to actual business processes is often missing.</p><p>This is where business impact becomes important.</p><blockquote>Where should AI make a measurable contribution? Which processes should change? What capabilities do employees need to make that change useful?</blockquote><p>Without that connection, AI risks remaining an additional layer of technology rather than changing how the company operates.</p><p><strong><em>3. Can people still judge the output?</em></strong></p><p>More AI does not remove the need for expertise.</p><blockquote>It makes expertise different.</blockquote><p>Even when AI takes over more tasks, companies still need people who can interpret results, verify them, and apply them responsibly.</p><p>For leaders thinking about Enterprise AI, this matters. Scaling AI cannot simply mean scaling usage. It also requires the knowledge to understand when an output is useful and when it is not.</p><p><strong><em>4. Does your AI position attract the people you need?</em></strong></p><p>This is the part I have been thinking about more recently.</p><p>Our AI positioning increasingly influences recruiting.</p><p>Good people are not only looking for a job. They want to work somewhere they can learn, shape things, and actually use new technology.</p><p>When candidates see that a company is not merely talking about AI but changing its own way of working with it, that becomes relevant.</p><blockquote>And it can create a useful cycle.</blockquote><p>A clear AI-first position attracts people who want to work with AI. Those people bring new knowledge and energy into the company. That, in turn, strengthens the company’s ability to work with AI.</p><p>For me, this is an important part of AI Leadership.</p><p>Technology alone does not determine whether a company stays ahead. The people you can attract, develop, and retain matter just as much.</p><h3><strong>We should be ambitious without pretending the transformation is finished</strong></h3><p>Marco compared the significance of AI with the Internet and argued that its impact could become even greater.</p><p>At the same time, we should be careful about short-term expectations.</p><p>There are already many experiments, use cases, and production applications. But relatively few companies have been fundamentally transformed by AI.</p><p>AI Transformation becomes meaningful when AI is consistently connected to processes, collaboration, and business models. Until then, a portfolio of pilots should not be confused with transformation.</p><p>For Swiss SMEs in particular, there is another practical dimension.</p><p>AI can help companies accomplish more with a limited workforce, make knowledge more accessible, and free capacity for work that creates more value.</p><p>That is a much more useful discussion than reducing AI to a debate about eliminating jobs.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CSyiB_jvuyaSG8OyyX3-EQ.jpeg" /><figcaption>Marco Denzler</figcaption></figure><h3><strong>The next constraint may be people, not technology</strong></h3><p>The technology will continue to move quickly.</p><p>We will see new tools, better models, and new approaches such as <em>Agentic AI</em>. Platforms and products will continue to evolve as well.</p><p>But companies cannot outsource the organizational part of the transformation to technology.</p><p>They still need people who are willing to learn.</p><p>They need expertise to evaluate what AI produces.</p><p>They need leaders who connect experimentation to actual business processes.</p><p>And they need an environment that talented people actively want to join.</p><p>So here is a practical question for leadership teams:</p><blockquote>If a strong candidate looked at how your company uses AI today, would they see a place where they could learn and shape the future of work, or simply a company that has bought some AI tools?</blockquote><p>If you’d like to continue the conversation, feel free to connect with me on LinkedIn: <a href="https://www.linkedin.com/in/balzzuerrer/">linkedin.com/in/balzzuerrer</a>.</p><p>As Group CEO of Online Group, I help organizations turn AI and digital transformation into measurable business outcomes. Learn more at <a href="https://www.online.ch/en/">online.ch</a></p><p>Interested in the full conversation? The second episode of «<strong>shift — The Signal in the Noise</strong>» is available in Swiss German on <a href="https://www.youtube.com/watch?v=4Fuosjj9x-4">YouTube </a>(with German subtitles) and <a href="https://open.spotify.com/episode/02Np1olGrbJ7trcIRpsnJc">Spotify</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*syUYahxK6ABNOmm0wpCWUw.jpeg" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5591f6a58aa7" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Powerful, Open, Cheap. Still Not Our AI Choice.]]></title>
            <link>https://medium.com/@balz.zuerrer/powerful-open-cheap-still-not-our-ai-choice-cd05e1af9bfe?source=rss-91cac6b5caa6------2</link>
            <guid isPermaLink="false">https://medium.com/p/cd05e1af9bfe</guid>
            <category><![CDATA[ai-transformation]]></category>
            <category><![CDATA[ai-first]]></category>
            <category><![CDATA[ai-leadership]]></category>
            <category><![CDATA[enterprise-ai]]></category>
            <category><![CDATA[agentic-ai]]></category>
            <dc:creator><![CDATA[Balz Zürrer]]></dc:creator>
            <pubDate>Fri, 14 Aug 2026 06:16:01 GMT</pubDate>
            <atom:updated>2026-08-14T06:16:01.355Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*UtrX0D7DAFl7a39XCLFh1Q.png" /><figcaption>Created with AI</figcaption></figure><p><strong>Chinese AI models are becoming powerful, open, and cost-efficient. For many companies, that makes them increasingly attractive. We still choose not to use them.</strong></p><p>With Kimi K3, Chinese company Moonshot AI has released an open-weight model that comes technically close to leading models from OpenAI, Anthropic, and Google.</p><p>In some areas, including coding and Agentic AI tasks, it matches or even outperforms them.</p><p>At the same time, it is openly available and comparatively inexpensive.</p><p>That is an attractive combination.</p><p>Yet we do not use Chinese AI models in our solutions.</p><p>The reason is not performance. It is sovereignty.</p><h3><strong>Data sovereignty is not model sovereignty</strong></h3><p>Chinese open-weight models are sometimes positioned as a sovereign alternative to American AI platforms.</p><p>The argument is not entirely wrong.</p><p>If a company runs a model itself or uses a European provider, it gains more control over its data and infrastructure.</p><p>That matters.</p><p>But data sovereignty and model sovereignty are two different things.</p><p>The model itself, its training data, and its fundamental characteristics remain Chinese in origin.</p><p>Running it on a European server does not turn it into a European AI system.</p><p>This distinction matters when companies make long-term decisions about Enterprise AI and their broader AI Transformation.</p><h3><strong>Models are shaped by more than technology</strong></h3><p>Powerful AI models are shaped by their training data and alignment.</p><p>As a result, they also reflect societal and political perspectives.</p><p>With Chinese models, this becomes visible in areas such as Taiwan, human rights, or freedom of expression.</p><p>These fundamental characteristics cannot simply be neutralized by adding another set of rules or a system prompt.</p><p>This is why I believe model selection should not be treated purely as an IT or cost decision.</p><p>When a company chooses an AI model, it is also deciding which technological origin, governance, and underlying values it is prepared to trust.</p><p>That makes model selection an AI Leadership decision, not just a technical one.</p><h3><strong>Sovereignty has at least three layers</strong></h3><p>For me, a useful way to evaluate AI sovereignty is to separate it into three dimensions:</p><ol><li><strong>Data sovereignty</strong><br> Where is the data processed, and who controls it?</li><li><strong>Operational sovereignty</strong><br> Where does the model run, and who controls the infrastructure?</li><li><strong>Model sovereignty</strong><br> Where does the model come from, and which governance and underlying values shaped it?</li></ol><p>Looking only at server location answers part of the question.</p><p>It does not answer all three.</p><p>A model can run entirely on European infrastructure while its origin and fundamental characteristics remain unchanged.</p><p>That distinction should be part of any serious Enterprise AI governance discussion.</p><h3><strong>Why we have made our choice</strong></h3><p>Our position is clear.</p><p>We do not use Chinese AI models in our solutions, even when they are technically strong and, in some cases, cheaper.</p><p>This is a conscious decision about which dependencies we are willing to accept.</p><p>For European companies, sovereignty should not stop at the data center.</p><p>Data matters.</p><p>Operations matter.</p><p>And the model itself matters.</p><p>When evaluating your next AI model, look at all three layers separately.</p><p><strong>Are you choosing only where your AI runs, or are you also consciously choosing what you are building on?</strong></p><p>If you’d like to continue the conversation, feel free to connect with me on LinkedIn: <a href="https://www.linkedin.com/in/balzzuerrer/">linkedin.com/in/balzzuerrer</a>.</p><p>As Group CEO of Online Group, I help organizations turn AI and digital transformation into measurable business outcomes. Learn more at <a href="https://www.online.ch/en/">online.ch</a></p><p>Interested in my podcast? «<strong>shift — The Signal in the Noise</strong>» is available in Swiss German on <a href="https://www.youtube.com/playlist?list=PLH70kG1NyibE">YouTube</a> (with German subtitles) and <a href="https://open.spotify.com/show/033PfO15H2YOis98RUhIme">Spotify</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*62xJ46s2dMQjBEk26VUokQ.jpeg" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=cd05e1af9bfe" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[In Search of the Signal]]></title>
            <link>https://medium.com/@balz.zuerrer/in-search-of-the-signal-40b0c5bd3ba5?source=rss-91cac6b5caa6------2</link>
            <guid isPermaLink="false">https://medium.com/p/40b0c5bd3ba5</guid>
            <category><![CDATA[ai-transformation]]></category>
            <category><![CDATA[enterprise-ai]]></category>
            <category><![CDATA[impact-business]]></category>
            <category><![CDATA[ai-leadership]]></category>
            <category><![CDATA[ai-first]]></category>
            <dc:creator><![CDATA[Balz Zürrer]]></dc:creator>
            <pubDate>Fri, 31 Jul 2026 12:26:54 GMT</pubDate>
            <atom:updated>2026-08-03T05:58:12.310Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*PRQ-35tziBXul5y6PkNgbA.png" /><figcaption>Created with AI</figcaption></figure><p><em>Enterprise AI is changing the way organizations operate. The harder question is how leaders can make sense of everything that comes with it.</em></p><p>Over the past few years, the conversation around Enterprise AI has accelerated dramatically.</p><p>Every week brings new models, new tools, new predictions, and new expectations.</p><p>Yet I believe most leaders are facing a different challenge.</p><p>The problem is not keeping up with technology.</p><p>The problem is knowing which developments will create lasting business impact and which are simply part of the noise.</p><p>That is why I started <a href="https://open.spotify.com/show/033PfO15H2YOis98RUhIme">«<strong>shift — The Signal in the Noise</strong>»</a>.</p><h3><strong>Why I Started a Podcast?</strong></h3><p>There is already no shortage of content about AI.</p><p>Articles, webinars, conferences, LinkedIn posts — there is more information available today than ever before.</p><p>What I often find missing is thoughtful conversation in Swiss German.</p><p>The most valuable lessons I have learned throughout my career have rarely come from presentations or trend reports. They have come from conversations with leaders who have faced uncertainty, made difficult decisions, learned from failure, and successfully led change inside their organizations.</p><p>Those are exactly the conversations I wanted to share.</p><p>In every episode, I speak with entrepreneurs, researchers, and executives who are actively shaping the future of business. Together, we explore AI Leadership, digital transformation, organizational change, and what it really takes to build companies that are ready for the future.</p><h3><strong>Looking Beyond Technology</strong></h3><p>Technology has never been the most interesting part of transformation for me.</p><blockquote>People are.</blockquote><p>Building an AI-first organization is not about deploying more technology.</p><p>It is about creating the leadership, culture, and decision-making capabilities that allow technology to generate real value over time.</p><p>That is why I believe successful AI Transformation is ultimately a leadership challenge rather than a technology project.</p><h3><strong>The First Conversation</strong></h3><p>To begin the series, I invited <a href="https://www.linkedin.com/in/petertsany/"><strong>Peter Thomas Sany</strong></a>, an internationally respected executive whose career spans more than four decades of leadership and transformation.</p><p>Peter has held CIO and CTO roles at organizations including Novartis, Deutsche Telekom, UBS, and Swiss Life. Today, he supports executives and organizations worldwide as an executive advisor, mentor, and keynote speaker.</p><p>There was no better person to start this journey with.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*WCluPV7LmU_vu2hxh5cpng.jpeg" /><figcaption>Peter Sany</figcaption></figure><h3><strong>Leadership in an AI Era</strong></h3><p>Our conversation goes far beyond technology.</p><p>We discuss leadership, lifelong learning, organizational change, and what it means to make decisions when certainty simply does not exist.</p><p>One message stood out throughout our discussion.</p><blockquote>Technology can accelerate change.</blockquote><p>But technology alone cannot transform an organization.</p><p>Successful AI Transformation depends on trust, curiosity, continuous learning, and leaders who help people navigate uncertainty instead of pretending to have all the answers.</p><p>As AI continues to evolve, the human side of leadership becomes even more important.</p><h3><strong>More Signal. Less Noise.</strong></h3><p>The name of the podcast reflects what I believe many executives experience every day.</p><blockquote>We are surrounded by more information than ever before.</blockquote><p>Finding clarity has become much harder.</p><p>With <a href="https://open.spotify.com/show/033PfO15H2YOis98RUhIme">«<em>shift — The Signal in the Noise</em>»</a>, my goal is not to predict the future or promote the latest technology.</p><p>It is to create conversations that help leaders think more clearly, challenge their assumptions, and better understand the forces shaping their organizations.</p><p>The podcast complements my LinkedIn newsletter of the same name, where I explore many of the same topics from a different perspective.</p><p>Together, both formats share one mission:</p><blockquote>To separate meaningful signals from temporary noise — and help leaders make better decisions in the age of AI.</blockquote><p>If you’d like to continue the conversation, feel free to connect with me on LinkedIn: <a href="https://www.linkedin.com/in/balzzuerrer/">linkedin.com/in/balzzuerrer</a>.</p><p>As Group CEO of Online Group, I help organizations turn AI and digital transformation into measurable business outcomes. Learn more at <a href="https://www.online.ch/en/">online.ch</a></p><p>Interested in the full conversation? The first episode of «<strong>shift — The Signal in the Noise</strong>» is available in Swiss German on <a href="https://www.youtube.com/watch?v=AorIteoredE&amp;list=PLH70kG1NyibE&amp;index=2">YouTube</a> (with German subtitles) and <a href="https://open.spotify.com/episode/1bvRttsHBWmjMNBd1QJzB4">Spotify</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*C7OHq6PbaxgwkUHRZA-g7g.jpeg" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=40b0c5bd3ba5" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Most Companies Don’t Need More AI]]></title>
            <link>https://medium.com/@balz.zuerrer/most-companies-dont-need-more-ai-d0dfe4462778?source=rss-91cac6b5caa6------2</link>
            <guid isPermaLink="false">https://medium.com/p/d0dfe4462778</guid>
            <category><![CDATA[impact-business]]></category>
            <category><![CDATA[ai-leadership]]></category>
            <category><![CDATA[ai-transformation]]></category>
            <category><![CDATA[ai-first]]></category>
            <category><![CDATA[enterprise-ai]]></category>
            <dc:creator><![CDATA[Balz Zürrer]]></dc:creator>
            <pubDate>Fri, 17 Jul 2026 14:04:09 GMT</pubDate>
            <atom:updated>2026-07-31T12:37:51.734Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bk_01jyuMlOsanX1jwtYLQ.png" /><figcaption>Created with AI</figcaption></figure><p>The AI Spine is an operating model, not another AI project<em>.</em></p><p>Most organizations don’t struggle with access to AI. They struggle with turning AI into measurable business outcomes. Employees have access to powerful tools, pilot projects are everywhere, yet enterprise-wide impact remains the exception rather than the rule.</p><p>A recent article in <em>MIT Sloan Management Review</em>, <em>Create Generative AI Value at Scale</em>, explains why. Based on a three-year study of 23 Swiss companies and 87 interviews with executives and business experts, the researchers found that only two companies successfully scaled AI across the enterprise. Although they developed their approaches independently, both had built what the authors call an AI Spine.</p><p>The conclusion is straightforward. Sustainable value from Enterprise AI does not come from isolated use cases. It comes from an organizational structure that embeds AI across core business processes and business units.</p><p>At Online, this reflects our own experience. After our strongest year in 2025, we continue to grow profitably while fundamentally changing how we operate. The common factor is not a particular AI application. It is the way AI has become part of our core business processes.</p><h3><strong>Start with Business Impact</strong></h3><p>Every AI initiative should contribute to a measurable business outcome: higher revenue, lower costs, better quality or faster execution. If none of these improve, the initiative may be interesting, but it is unlikely to create lasting Business Impact.</p><p>This principle guides every investment we make. We do not invest in impressive demonstrations. We invest in business results. Today we see efficiency improvements of 5 to 15% across the organization, while selected core processes have the potential to improve by more than 50%. At the same time, our position as an AI partner on Microsoft Cloud has become significantly stronger.</p><h3><strong>Broad adoption and focused transformation</strong></h3><p>Our implementation follows what we describe as a T-Shape model. The horizontal dimension is broad adoption. Every employee received access to ChatGPT Team and Microsoft Copilot through Microsoft 365, both providing the level of data protection required for secure work with business information.</p><p>This gave employees the opportunity to discover where AI creates value in their daily work. Beyond measurable productivity gains of 5 to 15%, it has also strengthened our attractiveness as an employer.</p><p>The vertical dimension focuses on transformation. We examined our core value creation processes and asked where AI could fundamentally change the way work is performed and where efficiency improvements above 50 percent were realistic. Two areas clearly stood out: software development and revenue management.</p><p>This combination of broad adoption and deep process redesign forms our approach to AI Transformation.</p><h3><strong>AI Leadership starts at the top</strong></h3><p>AI is not another IT project. It changes how a company creates value, which makes it a leadership responsibility.</p><p>At Online, AI Leadership starts with the CEO. Leaders need to work with AI themselves, share practical experience and demonstrate why their approach delivers results. That is why I deliberately discuss this topic in forums such as the HSG Alumni AI Club and at conferences.</p><p>The same leadership pattern is visible in other organizations. Andreas Berger, CEO of Swiss Re, recently stated in <em>NZZ</em>:</p><p>“With AI agents, we have reduced processing times from three weeks to as little as one day. Productivity improvements reach up to 80%.”</p><p>Technology alone is not enough. Successful implementation requires business owners who define value, domain experts who contribute process knowledge, IT teams responsible for integration, security and scalability, and AI Engineers who translate these requirements into solutions that work in daily operations.</p><p>This creates a shared digital backbone rather than another isolated application. That backbone is the AI Spine.</p><h3><strong>Transforming the core processes</strong></h3><p>Software development experienced the fastest change. Since the first quarter of 2025, our Software Engineers spend very little time writing traditional code. Instead, they orchestrate AI agents that support coding, security, penetration testing and quality reviews.</p><p>This is how we use specialized AI agents across the software development process. It reflects the direction widely described as Agentic AI, where multiple specialized agents work together to complete different tasks. We currently observe efficiency improvements of up to 60%.</p><p>As Michal Cisło, AI Engineer at Online, describes it:</p><blockquote>“I haven’t written a single line of code for months.”</blockquote><p>Revenue management is evolving differently. Here we are orchestrating digital marketing and sales activities with AI agents. Our Lead Agent identifies promising sales opportunities by combining multiple market signals. When a company repeatedly visits our website, several employees attend webinars and the company is recruiting an AI Business Expert, the system recognizes this combination as a strong buying signal.</p><p>AI also supports customer management. Agents identify customers who receive too little attention and remind account managers by email to reconnect. During periods of high workload, this helps ensure that important customer relationships receive the attention they need.</p><p>We are also applying AI to proposal creation. Progress in this area has been slower than expected because our proposals are not yet sufficiently standardized. Our experience shows that proposal generation becomes more challenging when the underlying process is not sufficiently standardized.</p><h3><strong>An operating model, not another AI project</strong></h3><p>The outcomes are measurable. 2025 was the strongest year in Online’s history, and we continue to grow profitably while strengthening our position as a leading provider of AI solutions on Microsoft Cloud.</p><p>Interestingly, demand is no longer our primary constraint. Our biggest challenge is growing our team quickly enough to deliver customer projects at the quality we expect. In one case, we even decided not to take on a new customer because our capacity would only have allowed us to start several months later.</p><p>For me, this is the most important insight from both the MIT Sloan research and our own experience.</p><p>The AI Spine is not a technology project. It is an organizational model that connects leadership, business ownership, technology and core processes around measurable business outcomes.</p><p>It also illustrates what an AI-first operating model looks like in practice.</p><p>A useful question remains for every leadership team:</p><blockquote><strong>Are your AI initiatives still a collection of isolated projects, or has Enterprise AI become part of the way your business operates?</strong></blockquote><p>I’m always interested in exchanging perspectives with people working through similar transformations. Let’s connect on LinkedIn: <a href="http://linkedin.com/in/balzzuerrer">linkedin.com/in/balzzuerrer</a>.</p><p>As Group CEO of Online Group, I work on organizational transformation, AI, and Microsoft Cloud initiatives in practice. More at <a href="http://online.ch/">online.ch</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/875/0*8DnBj1bf3_c_xDDi.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d0dfe4462778" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>