Organizational Structure and Design

Following up on my recent post about centralizing vs. decentralizing AI governance, I came across a Gartner report that sheds light on how organizations are actually structuring themselves for success, which matches my experience when working with top teams on strategy and AI. The Gartner study focuses on AI leaders whose organizations have already deployed at least one AI use case in production (n=251 for the structure data), so we're looking at insights from teams already in the game. Here are a few of my key takeaways from the research: 1. Your Challenges Evolve With Your Maturity The report clearly shows that low- and high-maturity organizations are solving completely different problems (Figure 2): - Low-maturity orgs are stuck at the starting line. Their top challenges are finding the right use cases (20%), governing generative AI use (12%), and funding initiatives (10%). - High-maturity orgs have different worries. Their problems are about scale and risk: security threats (19%) and data availability (11%). This shift from "What should we do?" to "How do we scale this securely and effectively?" is a critical milestone. 2. "Fit-for-Purpose" Beats "One-Size-Fits-All" This is the most compelling data, tying back to our governance discussion. The report analyzed which organizational structures (Centralized, Hybrid, Decentralized) had a significant positive or negative impact on the capabilities of high-maturity organizations (Figure 3). - Pure Decentralization showed a significant negative impact on critical areas like AI development teams, AI data, and AI ideation/prioritization. - Centralized models showed a strong positive impact for AI data, AI development teams, and AI application development. This makes sense for pooling scarce talent and ensuring standards. - Hybrid models were the winning strategy for areas like AI portfolio management, AI funding, and AI vendor management, balancing central oversight with business-unit autonomy. The key finding? High-maturity organizations don't default to one model. They design "fit-for-purpose" structures, centralizing what's high-risk or scarce (like data and core development) while using hybrid models for strategy and funding. How is your organization balancing the need for centralized standards with decentralized innovation? #AI #AIGovernance #DigitalTransformation #Leadership #GenAI #Gartner #DataStrategy #AIStrategy #OrgDesign

Most GRC teams are organised by accident, not design. After conversations with 150+ GRC leaders, I've mapped the exact decision framework that determines which team topology actually works for your current organisational reality. Tomorrow's GRC Engineer newsletter covers: 🏗️ The Three Architecture Patterns: • When centralized monoliths work (and when they become bottlenecks) • Why distributed chaos happens and how to avoid it • The hybrid platform model that scales with complexity ⚖️ The Decision Framework: • How your reporting structure (CISO vs. Legal vs. Risk) constrains your options • Why program drivers (compliance vs. risk vs. control coverage) matter more than team size • The hidden impact of Three Lines of Defense on organizational design 🎯 Organizational Reality Checks: • Why fewer than 3 GRC FTEs makes distributed models impossible • How compliance-driven vs. risk-driven programs need different architectures • When strict independence requirements kill embedding strategies 🚀 The Platform Pattern Deep-Dive: • Central services that function as "GRC operating system" • Embedded specialists as domain experts with dual reporting • Interface design that prevents coordination chaos 📋 6-Month Implementation Roadmap: • Phase 1: Platform foundation (policies, tools, protocols) • Phase 2: Strategic embedding in high-value teams • Phase 3: Scale and optimize with feedback loops 📊 Success Metrics & Anti-Patterns: • How to measure response time improvements and audit prep efficiency • Avoiding "Shadow GRC" and "Coordination Hell" • When to pivot between patterns as you scale The difference between GRC teams that scale and those that break isn't their tool, automation percentage or # of frameworks, it's their organisational architecture. Most teams think they're choosing between centralized control or distributed speed. The reality is more nuanced, and the stakes are higher than you think. If you think legacy is only your tech stack, think again. Subscribe to get the full breakdown tomorrow: https://lnkd.in/eFV3Yxi4 What's your biggest challenge with GRC team structure right now? Big shoutout to SecurityScorecard for sponsoring this week's issue. #GRCEngineering #TeamTopologies #GRCArchitecture

By 2030, 90% of data analysts will be decentralized. If they haven't been replaced by AI already, by then. Every week I speak to great data leaders and there is a clear pattern: The most successful and happy data leaders all decentralize to some extent. And it makes sense. Analysts who sit in a central ivory tower: - lack context on how the business works - are too far away from decision making - have no skin in the game Some arguments of the pro-centralization group: - Consistent and high talent bar for hiring analysts - Better growth opportunities for analysts in a central org - Consistency of methodology - More control over team culture Those are all fantastic arguments. But all these benefits can be achieved in hub-and-spoke orgs with the right management systems and the right data platform. My view: → Start out centralized and build a strong foundation → Rapidly turn business stakeholders into super users → Decentralize, starting from downstream roles (super users → analysts → analytics engineers) → Monitor ratios between roles per biz domain (e.g. analysts to analytics engineers) → Decentralization is not black and white, there are many flavors → Cross-functional teams can work really well if there is a strong culture of goal setting and alignment The rise of AI will further accelerate decentralization. One big challenge remains: None of this works without a strong data foundation. Neither with AI nor without. What’s your favorite setup? P.S.: Follow me Sebastian Hewing 🚀 for tips and strategies on creating more business impact with your data team. ♻️ Reshare to anyone who needs to hear this.

Centralized or Decentralized? In today's complex IT environment, global organizations grapple with choosing the best IT Service Management structure. Let's explore both centralized and decentralized ITSM frameworks, considering their impacts from a leadership and customer perspective. Centralized ITSM Pros: - Unified Strategy: Ensures a consistent approach to service management across all business units. - Cost Efficiency: Reduces duplicate efforts and leverages economies of scale. Centralized ITSM Cons: - Slower Response Times: Central control may delay localized decision-making. - One-Size-Fits-All: May not address specific regional requirements effectively. Decentralized ITSM Pros: - Agility: Local teams can respond swiftly to their specific needs and challenges. - Customized Service: Tailors services to regional customer demands and cultural nuances. Decentralized ITSM Cons: - Higher Costs: Potential for duplication of roles and technology across regions. - Inconsistency: Varying service levels and experiences can confuse customers and dilute brand integrity. Measuring Success: To assess the effectiveness of your ITSM structure, consider these metrics: 1. Customer Satisfaction Scores: Direct feedback on service quality and responsiveness. 2. Operational Efficiency: Analysis of incident resolution times and resource utilization. 3. Cost Metrics: Monitoring of expenditure against service delivery improvements. Organizations must weigh these factors carefully, aligning their ITSM structure with overarching business goals and customer expectations. Questions to Ponder: - Are we sacrificing quality for consistency with a centralized approach? - Could decentralized ITSM offer a competitive edge by being closer to our users? - How do we balance global standards with local excellence? What's your take on finding the right balance in ITSM frameworks? Reference: • ITSM People, "The Great Debate: Centralised vs. Decentralised ITSM for Global Organisations" • Picture Value Insights Website Cheers! #itil #itsecurity

Centralization (tops down) over decentralization (bottoms up) will make you go faster with your AI strategy and execution inside your company. and honestly, I think a lot of companies are going to have a spike in AI adoption, but hit a wall in AI maturity from taking the decentralized (bottoms up) approach first. The decentralized approach sounds great: - give everyone access - train everyone - encourage experimentation - let teams build And to be fair, this looks and feels like you're getting sh*t done. But at some point, the companies that actually scale AI successfully are going to need stronger top-down strategy, governance, and prioritization. Otherwise you end up with: - fragmented use cases - duplicate agents - conflicting outputs - different teams pulling different data sources - no prioritization framework - hundreds of AI experiments that never scale Here is why I lean toward centralization long term: Biggest ROI first The biggest wins come from documenting use cases across the business and prioritizing the most manual, repetitive, high-impact workflows, not just automating whatever annoys people individually. Trusted data matters If every team is pulling different inputs into agents, eventually the organization starts debating whose AI answer is “right.” Or people start using their own custom data sets for their agents and outputs never quite look the same. Governance becomes unavoidable Building agents honestly is not the hard part. Maintaining them compliantly at scale is. If everyone is building agents, who owns: - maintenance - retraining - governance - roadmap prioritization - feature improvements (There’s also a real buy vs. build conversation buried inside this). Standardization is how you scale If every team is building custom workflows and custom agents, it eventually becomes incredibly difficult to scale and consolidate across the organization. Quality comes from consistent outputs. Consistent outputs come from standardized processes. Scale comes from standardization. Build AI on top of that. Not everyone needs to build AI (probably my most controversial take) Everyone should absolutely become AI-enabled. Not everyone needs to become an AI engineer. At the end of the day, people still need to drive business outcomes. AI can become a distraction if teams become more focused on building the next cool agent than actually doing the job. Sellers just need to sell. Centralization vs. decentralization... curious as to where you all land on this.

The question I keep hearing about AI implementation is "should we centralize or decentralize?" Lots of smart folks like Kyle Norton are testing these models and sharing informed opinions and some pretty outsized results. I've been thinking a lot about this as I build out the marketing organization at Sequel.io and have opinions, but wanted to do actually research this and see what the data says. What I discovered is that there are pros and cons to each approach and the best solution probably lies somewhere in the middle. ➡️ On one hand, BCG found that decentralized AI models capture about half the ROI of centralized ones, mostly because the solutions that teams build don't scale. ➡️ On the other hand, McKinsey found that centralized AI teams become bottlenecks, starved of the local context they need to actually build useful things. Both findings are correct. That's probably why the companies seeing the most success are generally the ones using what's called the "federated model" which involves centralizing AI governance and decentralizing execution. The idea is that when you give people a stable, governed AI platform and a clear set of rules for how it should be used, then get out of their way, you'll see solutions that are better, developed faster, and used more broadly across the organization because the people feeling the daily pain are precisely the ones involved in building the solution. I went deeper on this in the latest issue of Code Meets Creed. You can read it here 👉 https://lnkd.in/ebgi9s9k I'd love to hear from other marketing leaders who have tested either of these models, or have strong opinions about them. What do you think the right approach is, and if you've implemented one of them, how is it going? #AI #marketing #kathleenhq

Every company that internationalizes has the same question: "Do we centralize marketing at HQ, or localize with teams on the ground?" It’s not an easy answer, but there’s a simple rule of thumb: - Centralize what's scalable and synergistic - Localize what requires proximity In practice that often means: 1. Performance marketing stays centralized (with local inputs). Google Ads and Meta are the same everywhere, so the operating system can be centralized: account structure, naming conventions, measurement, bidding logic, reporting, experimentation cadence. What still needs local input: creative, language, offers, landing pages, and market-specific insights. 2. Offline and relationship-driven channels stay localized. TV is negotiated country by country. Out-of-home is hyper local. Same for influencer activations, community, partnerships, and anything that depends on local networks, speed, and cultural nuance. That’s the rule of thumb. But a few variables change the decision. Here are 6 that matter most: 1. Capital: Localization is expensive. If you don’t have the budget, you’ll centralize by necessity. 2. Talent density: Centralizing across markets requires international talent (and often multiple languages) in one place. Some hubs make that much easier. 3. Cost vs attractiveness: Some locations attract global talent, but cost structures can make centralized teams less efficient. 4. Business model: SaaS and platform models often need less local marketing presence than logistics-heavy or service-heavy businesses. 5. Growth motion: Marketing-led and product-led can run more centrally. Sales-led needs native-speaking coverage and local enablement, which pushes you toward localization. 6. Revenue concentration: When one market becomes large enough, it usually earns dedicated ownership. ___ Most companies don't think about this when expanding to their first market. They run everything from HQ. It's when they market 3, 4, 5 that the question becomes real. TL:DR Start with the rule of thumb, then stress-test it against the variables. P.S. There’s a set of dimensions companies often underestimate when choosing their operating model. Find them in the next issue of Growth Beyond Reach.

🌐Organizational Archetypes for Generative AI: Centralized vs. Decentralized In the rapidly evolving landscape of generative AI, institutions are spearheading innovation. Based on my own experiences managing strategic projects and engaging in team dynamics discussions and recent discussions with senior leaders, let's delve into these archetypes and their implications for the future of generative AI organizational design and services. 🤝This analysis is inspired by excellent research from McKinsey, focusing on financial institutions, but it is broadly applicable across various industries and image is below. I’ll save the discussion on the build vs. buy equation for another post. 1️⃣. Highly Centralized: Description: All decision-making and execution are managed centrally. Pros: Ensures consistency, streamlined processes, and strategic alignment. Cons: May stifle creativity and slower to adapt to local needs. Results: 70% of institutions adopting this model are moving into production phases. 2️⃣ Centrally Led, Business Unit Executed: Description: Central strategy with execution handled by individual business units. Pros: Balances strategic oversight with operational flexibility. Cons: Potential for misalignment between strategy and execution. Results: 50% of institutions are successfully starting production. 3️⃣ Business Unit Led, Centrally Supported: Description: Business units take the lead with support from a central AI team. Pros: Promotes innovation and agility within units. Cons: Challenges in maintaining unified strategic direction. Results: 50% have transitioned to production stages. 4️⃣ Highly Decentralized: Description: Independent business units manage both strategy and execution. Pros: High adaptability and innovation tailored to specific markets. Cons: Risks of fragmentation and inconsistency. Results: 30% are advancing to production. The choice between centralized and decentralized models hinges on organizational size, regional diversity, and specific use cases. 🎯 From my two decades in digital transformation and engineering leadership, I've found that a balance between these extremes, emphasizing cross-collaboration, transparent communication, and a strong central vision while empowering local units, yields the best results. For initiatives, centralized pilot projects can be started considering regional needs and then slowly moved towards a decentralized model. Key Considerations: 1️⃣Global Regions: Adapt to regional market dynamics and regulatory environments. 2️⃣ Team Dynamics: Encourage collaboration between central and local teams. 3️⃣ Technology Stack: Invest in robust infrastructure to support both centralized oversight and decentralized innovation. #GenerativeAI #DigitalTransformation #AILeadership #Innovation #TechStrategy #TeamDynamics #GlobalStrategy #FutureOfWork Sanjay Macwan Sumeet Chabria Tony Thomas Vikas Wali Sanjay Mazumder Shilpi Sharma Johannes Wolvius Shahram Ebadollahi Thomas A. Stewart

“What QMS structure should we build now… so we’re not rebuilding it in two years?” That’s the question a founder asked me last month, right after landing their Series A and preparing to expand into two new markets. I’ve heard variations of it from at least six companies in the past quarter alone, which tells me this isn’t just a tactical choice anymore, it’s strategic. As companies grow in new markets, new manufacturing partners, and new product variations, the question of centralized vs. decentralized QMS shows up sooner or later. And there’s no one-size-fits-all answer. Here’s what I've seen work (and backfire) across the board: 🛡️ Centralized QMS ✔ Great for small-to-mid-size teams or companies expanding under a single quality umbrella. ✔ Consistency, audit readiness, and precise document control. ❌ Can be slow, inflexible, and frustrating for local teams who need speed and nuance. 🌍 Decentralized QMS ✔ Gives full autonomy to each site, which works for large multinationals or those with distinct product families. ❌ The flip side is duplication, inconsistent compliance, and fragmented oversight. 🔗 Hybrid QMS (the model I increasingly recommend) ✔ Combines corporate-level SOPs with region-specific adaptations. ✔ Allows for shared ownership, but ensures traceability and alignment during audits. ✔ Particularly helpful when simultaneously scaling across the U.S., EU, and APAC. I often help clients transition from decentralized to hybrid as their operations mature, or from centralized to hybrid as they expand globally and need more localized agility. In my experience, the question isn't just “Which model should we use?” It’s: “What level of control, flexibility, and clarity do we need today and 18 months from now?” #QualityManagementSystem #MedTech #RegulatoryCompliance #ScalableSystems #GlobalExpansion #Elexes #MedicalDevices

“Just make the call from the top” sounds efficient…until it isn’t. Early in my IT career, I thought strong governance meant central control. More alignment, more consistency, fewer mistakes. But I’ve since learned that centralized governance works only when a leader knows where to draw the line. You need a hybrid model: centralized where it matters, decentralized where it counts. With hybrid models, leaders say what needs to be built - but the “how” comes from the product team. They have the freedom to make decisions in real time, based on real user feedback. In practice, the best IT governance I’ve seen has three things: • Clear ownership of responsibilities • Alignment between business goals and tech execution • Documented decisions that avoid confusion down the line You keep architecture, security, and budget decisions at the org-wide level. Those impact everything. But execution belongs closer to the front lines because that’s where the context lives. That’s how companies move fast and build the right thing. For example, my team recently helped a homebuilder digitize the entire post-sale warranty experience AC units, appliances, everything - into a searchable, shareable digital folder. If you’ve ever been stuck reworking something because someone too far removed made a call based on guesswork, you know what I mean. Clarity without bureaucracy means governance that accelerates, not obstructs.