Distributed Decision-Making Processes

🤔 Weekend Reflections 👉 As we head into the #AIActionSummit, the idea of creating a CERN for AI—both in Europe and beyond—continues to gain momentum. This call has been further amplified by SoftBank's recent $500 billion investment announcement in the US and the release of Deepseek in China 🌍. 🤔 But is a centralized, CERN-style model the only way forward for sustained, responsible innovation in AI? 👉 In my piece for Frontiers Policy Labs, I proposed a different path: a polycentric, distributed approach to AI and science. This model addresses three key challenges in the current AI ecosystem: 1️⃣ Access to computational resources 💻 2️⃣ Access to high-quality data 📊 3️��� Access to purposeful AI modeling 🤖 🔗 Read my full article here: https://lnkd.in/ezXxaX_Z 👉 The same rationale can be applied to AI governance, much like the distributed internet governance model I proposed earlier. 🤔 Distributed governance offers a more resilient, flexible, and inclusive framework with several key advantages: ✅ Facilitates cooperation among existing and emerging actors without the need for new bureaucratic structures. It encourages decentralized dialogue on key issues, fostering more flexible and creative solutions to emerging issues and applications than a top-down, centralized system. ✅ Acts as a “routing” function, enabling interoperability and collaboration by adopting shared standards and common ontologies. This approach empowers dispersed actors to contribute innovative solutions, shifting decision-making power to communities and experts who might otherwise be excluded. ✅ Promotes information-sharing and evidence-based decision-making. Distributed governance networks prioritize data-driven approaches, allowing stakeholders to accurately evaluate the effectiveness of governance initiatives across different regions and contexts. ✅ Allows for both granularity (localization) and scale (globalization). Issue- or expert-based organizing principles help coordinate decisions at the local, national, regional, and global levels. This ensures local actors are included in global conversations and prevents issues from escalating unnecessarily (This will also be discussed on Tuesday at our event on Aligning Local and Global AI Governance - See https://lnkd.in/eb8xfJh9). Q How to design AI governance—not as a monolithic institution, but as a dynamic, interconnected network of nodes working toward a common good? 🔗 Read my paper: A Distributed Model for Internet Governance (and eager to hear how it may apply to similar challenges of AI governance): https://lnkd.in/ejyUtset #AIActionSummit #OpenScience #DistributedGovernance #AIInnovation #Collaboration #PolycentricAI #AIgovernance #Deepseek #CERN

🎉 Excited to share that our paper with Luke Rhee, "The Power and Limits of Distributed Representations in Strategic Decision Making," has just been published in Strategy Science! We tackle a fundamental organizational challenge: How can companies make better decisions by combining the partial insights of multiple specialists? Think of it like landing a plane—neither pilot nor copilot has complete information, but together they succeed. We call these collective cognitive models distributed representations, and we develop a formal theory of when they help or hinder performance. Key takeaways for designing decision processes: ⚖️ Averaging delivers robust performance across most settings—a reliable default choice 🤝 Unanimity protects against errors when good opportunities are rare and managers lack experience 👤 Specialists excel only when a single factor dominates (e.g., star power for blockbusters) 🎯 Experienced generalists can outperform all approaches—but they're extremely scarce in practice 💡 The core insight: There's no universally "best" structure. Effectiveness depends critically on the three-way interaction between individual expertise, aggregation method, and environmental conditions (complexity, uncertainty, opportunity abundance). Our framework extends Brunswik's lens model and introduces "decision boundaries" from machine learning—bridging individual cognition and organizational structure, two research streams that have evolved separately for 60+ years. This feels particularly relevant as organizations increasingly integrate AI into decision-making. Understanding human-AI distributed representations will be essential for designing effective hybrid systems. 📄 The paper is open access. Read it here: https://lnkd.in/gvETSmTm #Strategy #OrganizationDesign #DecisionMaking #StrategyScience #AI

I used to think leading a distributed team meant finding the right project management tools. Turns out, the tool that mattered most was something simpler: showing my work. When we were building a team of 180+ across the US and India during COVID, I noticed that the decisions that landed well weren't necessarily the ones that went "right" in hindsight. They were the ones where people saw how we got there. A product pivot that made sense in California looked reckless in Chennai - until they understood what customer data we were looking at. A budget reallocation that seemed obvious to leadership felt arbitrary to the team - until we walked through the trade-offs we considered. So we started operating like we were working in public, even internally. Before rolling out a major change, we'd write up the alternatives we rejected and why. We'd share the customer feedback or market signal that pushed us in this direction. We'd even admit what we were uncertain about. And we'd play actual recordings of customer and prospect calls in our all-hands meetings, including the messy conversations where people reacted to our proposed solutions in ways we didn't expect. Of course, some decisions still came down to a judgment call. We didn't always have consensus. But everyone could trace the logic and apply the same framework to their own decisions. To my surprise, this approach didn't slow us down. It distributed decision-making authority faster than we would have otherwise. When your team understands how you think, they don't need to wait for you to weigh in on everything. We don't do as much of this explicit writing and sharing today since we’re back together in person, but it came back to me recently when Vidhya Madhusudan mentioned during her 5-year milestone how impactful those call playbacks were. It reminded me that transparency builds something more valuable than alignment: a team that thinks like owners.

Every leadership team today claims they want agility. Very few are willing to redesign how decisions are made. Artificial Neural Networks learn by distributing intelligence across many connected nodes. No single node holds the entire truth. The power comes from the network learning together. The same pattern is now reshaping high-performing organisations. What we’re seeing in the field is clear: ● McKinsey’s 2024 Org Diagnostic found that companies with distributed decision-making outperform peers by up to 33% on speed and nearly 20% on innovation outcomes. ● Teams with autonomy resolve customer issues 40–60% faster, based on data from Bain’s Frontline of the Future study. ● Gartner reports that organisations relying on “centralised escalation chains” lose up to 27% productivity during periods of volatility. ● In India’s digital ecosystem, firms that decentralised operational decisions during 2023–24 scaled new product rollouts 1.7x faster. The shift is operational. In neural networks, weights determine how strongly a signal is amplified. Bias sets the threshold at which a node acts. Training is simply a feedback loop that adjusts these two levers of weights and bias, to improve outcomes. Organisations work the same way. Your “weights” are incentives, information flows, and decision rights. Your “bias” is leadership intent, risk appetite, and what you reward or tolerate. For decentralised intelligence to work, leaders need to build three things into how their organisations actually run: ● 𝗙𝗮𝘀𝘁𝗲𝗿 𝘀𝗶𝗴𝗻𝗮𝗹 𝗱𝗲𝘁𝗲𝗰𝘁𝗶𝗼𝗻 Distributed teams should be able to flag and act on weak signals early, especially when cycles tighten, customer expectations change weekly, and regulatory shifts arrive without warning. ● 𝗕𝗲𝘁𝘁𝗲𝗿 𝗰𝗼𝗻𝘁𝗲𝘅𝘁𝘂𝗮𝗹 𝗱𝗲𝗰𝗶𝘀𝗶𝗼𝗻𝘀 Frontline teams hold contextual knowledge that dashboards don’t. Leaders need to give them clear boundaries, data access, and the authority to decide. ● 𝗦𝗰𝗮𝗹𝗮𝗯𝗹𝗲 𝗹𝗲𝗮𝗿𝗻𝗶𝗻𝗴 𝗹𝗼𝗼𝗽𝘀 ANN models get better as they train by ‘fine-tuning’ weights and bias. Organisations that mirror this with structured learning loops and post-decision reviews see measurable uplift. The leadership challenge is simple: you cannot ask for agility while holding on to control. The question that senior leaders must now confront is: Are we building networks that learn, and fine-tune or structures that slow the learning itself down? #AI #Leadership

You're the bottleneck you once complained about. It's not control. It's unclear decision rights. Technical experts who've risen to leadership unconsciously create approval bottlenecks. Not because they don't trust their teams. But because they've built their identity on technical excellence. The pattern plays out in three costly ways: 1️⃣ Decision bottlenecks Projects stall while capable teams wait for your review. 2️⃣ Responsibility avoidance Teams stop fully owning problems that you'll review anyway. 3️⃣ Trust deficit Your approvals signal: "I don't quite trust your judgment." When I was struggling with letting go of the decision making, I tracked my decisions for two weeks and discovered: ↳ 75% could be made at lower levels ↳ My review added 2 days per decision ↳ <10% needed my technical expertise The problem isn't control-seeking behavior. It's unclear decision rights. Try this simple exercise with your leadership team: ✓ Write down every decision you reviewed last week ✓ For each, answer: → What's the worst possible outcome if someone else decided? → Is that outcome reversible? → Does reviewing this grow capability in others? ✓ Create 3 clear decision categories: → You Inform (team decides, just keeps you updated) → You Decide (true technical expertise needed) → You Approve (quick check for alignment) Your technical expertise got you here. But your ability to distribute decision-making will determine your success as a leader. P.S. If this resonated with you, share it with your network. ♻️ P.P.S. Many leaders face this challenge but few discuss it openly.

To successfully innovate at scale, modern organizations must evolve from the traditional top-down command structure to a more distributed leadership model, giving teams real authority to act, adapt, and innovate. A recent video from MIT Sloan Management Review explores how leaders at all levels can be empowered with real decision-making impact. When authority moves closer to the work, teams respond faster, learn in real time, and take greater ownership. Several shifts are necessary to make distributed leadership truly effective in practice: ⭐ A mindset shift around authority. Empowerment means trusting teams with decision rights, not just assigning tasks. ⭐ Structures that support autonomy. Clear priorities and guardrails allow teams to move quickly without waiting for approvals. ⭐ Capability over control. Teams build judgment by making decisions, seeing results, and adjusting course. ⭐ Readiness for ambiguity. Organizations that normalize uncertainty are better equipped to operate in more dynamic, complex environments. The real unlock comes when decision rights, accountability, and context are clearly aligned. That clarity allows teams to move with speed, agility and excellence to deliver meaningful results. Where could a shift in decision-making power create an impact in your organization? Learn more: https://lnkd.in/gEyiDgak

How do teams *actually* make good decisions under pressure? Not how do they say they do it, but how do decisions really get made when time, information, and certainty are limited? Dr. Mark Ramzy (from REBEL EM) and I explored this in a recent episode of The Emergency Mind Podcast, focusing on how teams make decisions in intensive care. In complex environments like the ICU, no single individual has enough information to make optimal decisions alone. Knowledge is distributed across nurses, pharmacists, respiratory therapists, surgeons, and physicians, each holding partial, local views of the system. Effective decision making depends on how those perspectives are integrated. That makes team performance a systems design problem as much as a leadership one. We talked about open versus closed ICUs, hierarchy versus more democratic models, and why rigid command-and-control structures often fail in complex systems. One insight that stayed with me: effective distributed decision making is not built through speeches or posters. It is built by practicing shared judgment on small, everyday decisions until the system learns how to self-organize when the stakes are high. Hope you enjoy. https://lnkd.in/gqpd9H-H #ComplexSystems #CriticalCare #DecisionMaking #AdaptiveLeadership