California Management Review

The Argument: Professional services firms have survived three technology-driven transformations: ERP implementation (1990s), web/mobile enablement (2000s), and SaaS/cloud platforms (2010s). Each wave changed what clients bought, how they paid, and what they received — but left the fundamental consulting model intact. Clients still purchased human expertise, measured in hours or FTEs. AI breaks this pattern. The fourth transformation inverts the services model entirely. Rather than software enabling consultants to work faster, expertise itself becomes software. I call this ‘Service as a Software’ — the encoding of domain judgment into autonomous systems that deliver outcomes directly, with humans supervising rather than executing. This article argues that the critical capability for this era is not AI engineering but Expertise Architecture: the systematic methodology for capturing domain judgment and encoding it into machine-executable reasoning. Firms that master this capability will capture disproportionate value. Those that treat AI as merely another accelerator for existing labor models will find themselves disrupted by focused entrants who start without legacy economics to protect. The Four-Era Framework Each technology wave transformed consulting economics in predictable ways: The Incumbent Response: Two Paths, Same Gap The industry’s largest players recognize the shift — but their responses reveal an unresolved strategic tension. On January 22, 2026, McKinsey and AWS launched the Amazon McKinsey Group (AMG), a joint venture explicitly designed around outcome-based pricing. The structure is notable: rather than billing for consultant hours, AMG ties fees to measurable transformation results on engagements exceeding $1 billion. This is a structural bet that the traditional labor model cannot survive Era 4. McKinsey is not adding AI to consulting; it is repositioning consulting around AI-enabled delivery. Yet even this bold move exposes the gap. AMG still depends on McKinsey consultants to interpret client context, design transformation roadmaps, and validate AI-generated recommendations. The ‘expertise layer’ remains human. The joint venture changes who bears performance risk, but it does not fundamentally change how expertise gets delivered. McKinsey has restructured the economics without yet encoding the judgment. Contrast this with Accenture’s approach. The firm announced $3 billion in AI investments and has built impressive technical capabilities — AI factories, proprietary tools, thousands of trained practitioners. But the underlying delivery model remains intact: consultants use AI to work faster, clients still pay for FTEs, and value is measured in hours saved rather than outcomes achieved. This is Era 3 optimization, not Era 4 transformation. AI augments the labor model; it does not invert it. Deloitte, EY, and others occupy similar positions — significant AI investment, genuine technical capability, but strategic ambiguity about whether AI is a tool for consultants or a replacement for consulting. The ambiguity is rational: these firms generate billions in labor revenue. Protecting that revenue while simultaneously enabling autonomous delivery creates organizational tension that no amount of investment resolves. The pattern across incumbents is consistent: recognition without resolution. They see the shift. They are investing heavily. But none has cracked the core problem — systematically encoding the domain expertise that makes their consultants valuable. Until they do, they remain vulnerable to focused entrants who start without legacy economics to protect. Where Service as a Software Already Works While incumbents navigate strategic tension, a new category of company demonstrates that ‘Service as a Software’ is not theoretical — it is already operating. Vertical AI companies target functional domains where decision patterns are bounded, judgment is repeatable, and outcomes are measurable. Rather than building general-purpose AI tools, they encode domain-specific expertise into autonomous systems that deliver results directly. The model works because these companies start with encoded expertise as their core asset, not as an enhancement to labor. Consider the pattern: A procurement AI platform doesn’t just surface contract anomalies for humans to review. It encodes the judgment that procurement professionals apply — what constitutes a meaningful price variance, which suppliers warrant scrutiny based on risk profile, when to escalate versus auto-approve. The system doesn’t accelerate human work; it executes the work with human oversight. In my own experience building an AI-native FinOps platform, I’ve observed how this plays out operationally. FinOps — the practice of managing cloud and technology spend — involves hundreds of decisions daily: Is this cost spike an anomaly or expected? Should this workload be rightsized? Does this variance warrant executive attention? Traditional approaches surface data and expect humans to decide. Our approach encodes the decision logic itself. We built what we call PRISM — a methodology that decomposes FinOps into five decision domains (Proactive monitoring, Resource optimization, Infrastructure management, Spend economics, and Management governance), each with explicit thresholds, escalation rules, and context-dependent reasoning. The AI doesn’t just detect a 15% spend increase; it knows that 15% in production environments during quarter-end is expected, while 15% in development environments on weekends warrants immediate investigation. The result: decisions that previously required analyst review now execute autonomously within defined risk parameters. Humans supervise exceptions rather than processing routine judgments. Time-to-action compresses from days to minutes. And critically, the value delivered is measurable in outcomes — cost avoided, efficiency gained — not hours worked. This pattern — bounded domain, encoded judgment, autonomous execution with human oversight — defines where ‘Service as a Software’ already works. The question for incumbents is whether they can replicate it before these focused players expand. Where It Fails: The Encoding Gap For every successful vertical AI deployment, there are dozens that stall — not because the AI doesn’t work, but because the expertise was never encoded. Early in our product development, we learned this lesson directly. We deployed anomaly detection to flag unexpected cost variances in cloud infrastructure. The model performed well technically — it identified patterns humans missed, processed data at scale, and surfaced potential issues in real time. By any AI benchmark, it succeeded. But in practice, it failed. The system flagged everything that deviated from baseline: a 12% increase in compute spend, a new storage allocation, a spike in data transfer costs. Within the first week, it generated over 200 alerts. Finance teams, already stretched thin, couldn’t process them. They had no way to distinguish signal from noise because the system didn’t encode what practitioners know — that a 12% increase during product launch is expected, that new storage allocations tied to approved projects aren’t anomalies, that data transfer spikes during backup windows are routine. Without encoded thresholds, context rules, and escalation logic, the AI created more work, not less. Alert fatigue set in within two weeks. Teams began ignoring notifications. By month two, the anomaly detection was effectively shelfware — technically operational, practically abandoned. We had automated data processing but not decision-making. This pattern repeats across enterprises. A legal team deploys AI to review contracts; without encoded risk tiers, lawyers still review 100% of outputs. A customer service team launches an AI assistant; without encoded resolution paths, 60% of queries escalate to humans. A finance team automates expense auditing; without context rules, 80% of flags are false positives. The failure mode is consistent: organizations deploy AI capability without encoding the judgment that makes capability useful. They automate the process layer — data ingestion, pattern recognition, alert generation — while leaving the expertise layer untouched. The result is augmentation at best, shelfware at worst. Why It Fails: The Missing Layer The failures above share a common root cause. Enterprises have invested decades in process frameworks — APQC taxonomies, BPMN models, SIPOC documentation. These capture workflow: what activities happen, in what sequence, with which roles. They do not capture judgment. Consider a common process step: ‘Review budget variance.’ A process model shows this as an activity box connected to a role. But what AI needs to automate this step is entirely different: When is a variance significant? 5%? 10%? Does it depend on the cost center? The time of year? The trend direction? These are judgment calls that exist in practitioners’ heads but not in any process documentation. This is the critical gap: AI can automate process. AI cannot automate judgment without encoding. I call this missing layer Expertise Architecture — the systematic encoding of domain judgment into machine-executable form: The strategic implication is clear: Process classification and process models are commoditized — everyone has them. The scarce, ownable capability is methodology for building the Expertise Architecture layer. Firms that build this layer first will define the category. What To Do: Guidance for Three Audiences For consulting firm leaders: Current IP libraries are necessary but insufficient for Era 4. The strategic question shifts from ‘how do we deploy AI tools’ to ‘how do we encode our expertise before others do.’ Protecting existing labor revenue will delay transformation; focused entrants face no such constraint. The McKinsey-AWS model shows one path — restructure economics around outcomes. But without encoded expertise, outcome-based pricing shifts risk without changing capability. For enterprise buyers: Evaluate vendors on expertise encoding methodology, not AI capability alone. Ask: ‘How do you capture and validate domain expertise in your AI systems?’ Demand transparency on human-AI task allocation. If a vendor’s AI still requires your team to review every output, you’re buying augmentation, not transformation. Expect and negotiate for outcome-based pricing as the standard — and verify the vendor has encoded enough judgment to deliver on it. For new entrants: Functional verticals (FinOps, procurement, revenue operations) offer clearer encoding targets than industry verticals. The bounded nature of these domains — repeatable decisions, measurable outcomes, defined thresholds — makes expertise encoding tractable. First-mover advantage accrues to firms that establish trust and governance frameworks. The window for category definition is open but will close within 3-5 years as incumbents resolve their strategic ambiguity. Conclusion The consulting industry has survived three technology transformations by adapting its delivery model while preserving its fundamental economics: clients pay for human expertise. AI breaks this pattern because it enables expertise itself to become software. The winners of this transformation will not be determined by AI capability — that is rapidly commoditizing. They will be determined by who solves the expertise encoding problem first. The firms that build Expertise Architecture — the systematic methodology for converting domain judgment into machine-executable reasoning — will capture disproportionate value. Those that treat AI as another tool for consultants will find themselves disrupted by focused players who started without legacy economics to protect. The shift from rate cards to outcomes is not a pricing change. It is an architectural change. And the architecture that matters most is not technical — it is the encoding of human expertise into systems that can act on it.

Imagine a Monday morning at a global corporation’s strategy meeting. The Chief Marketing Officer is pitching a bold new creative campaign. The Head of Engineering counters with product feasibility concerns. The CFO, focused on risk, winces at the projected costs. There is tension in the air — a classic case of functional diversity at work. Different backgrounds and areas of expertise are colliding. In a small company, this same mix of perspectives might spark an innovative pivot on the spot. But in a mega-corporation, it often leads to crossed wires and frustration. Why? Can leaders harness these differences for better performance, rather than letting them hinder progress? We explore these questions, drawing on our new research1. Key Takeaways for Busy Executives Not All Diversity Is Visible: Functional diversity captures the variety of occupations held by members in a team (finance, marketing, operations, etc.). While functional diversity is less visible than other forms of diversity such as race or gender, it can be a powerful driver of better decisions and innovation. Different functional backgrounds reflect distinct experiences and mental models, which enlarges a team’s collective skill set and perspective. Big Benefits for Small Firms: New evidence shows that functional diversity in the top management team has a strong positive impact on performance in smaller organizations. In startups or mid-sized firms, a mix of expertise may help spot opportunities and problems faster, boosting agility and results. Large Firms Face a Diversity Paradox: In giant companies, simply having diverse expertise on the top management team won’t automatically yield results. Complex hierarchies, siloed divisions, and communication breakdowns can smother the potential gains of diversity. The Integration Fix — Experience Matters: A long-tenured, well-knit top management team can turn functional diversity from friction to fuel. When executives have years of shared experience, they develop trust, a common language, and better ways to integrate their different perspectives. Rethinking the Business Case: High-profile reports claiming diversity causes higher profits (like McKinsey’s oft-cited studies) were correlational and have been critiqued for overstating conclusions. Our research — covering 4,500 firms and 32,000 executives over nearly two decades — used careful controls to better isolate causation. Top management team functional diversity can drive organizational performance, especially under the right conditions (smaller organizations or larger organizations with high team integration). The Overlooked Side of Diversity: Why Functional Backgrounds Matter “Diversity” often refers to demographics like gender, ethnicity, or nationality. These matter, but another powerful form is functional (or occupational) diversity: differences in professional backgrounds, such as a sales veteran, supply chain specialist, finance expert, and tech specialist. Each brings distinct mental models, vocabularies, and problem-solving approaches shaped by their training and career paths — creating a true “diversity of thought.” Functional diversity is a double-edged sword. Positively, teams with varied expertise approach problems from multiple angles, generating more ideas and preventing groupthink. A marketing executive proposes a bold customer-focused innovation, an operations leader refines it for efficiency, and a finance expert ensures profitability — yielding better overall decisions. However, knowledge differences create friction. Functions develop their own subcultures, with unique languages, priorities, and even stereotypes (e.g., storied tensions between finance vs. marketing, or R&D vs. sales). Studies have long shown mixed results on whether functionally diverse teams outperform homogeneous ones. The Categorization-Elaboration Model highlights why: while diverse groups can benefit from bringing together different perspectives and information, they also face social “us versus them” divisions that can make collaboration and integration difficult. In sum, unlocking diversity’s benefits requires effective cross-functional communication and collaboration, which is not automatic and harder as organizations grow larger. A Quick Example — The Big Pharmaceutical Dilemma  A drug development team: excited chemists, cautious clinicians, story-focused marketers, and compliance-wary regulators. Each perspective is essential. Good collaboration requires respecting expertise and finding alignment to build a better, safer, more-marketable drug. Poor communication (more common in big pharmaceutical organizations) will stall or sink the project. Functional diversity is only an asset when teams know how to harness it. Small Company, Big Advantage: Why Diversity Shines in Lean Organizations So, when does functional diversity pay off the most? According to our recent large-scale study of thousands of firms, we found that top-management team functional diversity led to higher profits and growth, but that this effect was limited to smaller and mid-sized companies. Little to no direct benefit was observed in the largest firms2. Why might that be the case? Smaller organizations are inherently more nimble. They often have less bureaucracy and more face-to-face interaction, which makes it easier for a diverse team to actually combine their knowledge. There’s seldom a maze of divisions to navigate or complex chain of command for approvals. An engineer, a designer, and a customer service expert all on a founding team can be like rocket fuel — each one sees different threats and opportunities, which they quickly act on together. Team research suggests people in smaller teams also tend to wear multiple hats, fostering appreciation of each other’s challenges. A startup COO might handle HR one day and supply chain the next, with these blurred boundaries teammates develop a shared context faster — the marketing person understands a few engineering constraints, the engineer appreciates the marketing angle, etc. Information thus flows with less friction, and the team can leverage every member’s expertise. As “members may need to fill multiple roles and have expertise in multiple areas,”3 functional diversity can be pragmatically managed in day-to-day interactions. Consider a small hypothetical tech startup. The sales-savvy CEO, code-focused CTO, ex-designer Product head, and banking-trained CFO argue frequently in weekly meetings — but resolve issues in real time, face-to-face. When planning a new feature, user-centric ideas get reality-checked by technical feasibility and budget constraints, leading to creative, balanced solutions (e.g., phased rollouts) in a single afternoon. In large companies, the same cross-functional negotiation often drags on for months via emails and global meetings, if it happens at all. The Diversity Paradox in Mega-Corporations: When Good Mixes Go Awry If functional diversity is so great, why don’t we see clear performance gains in big corporations? The answer lies in what we might call the diversity paradox of large organizations: the conditions that emerge with scale — size, complexity, and hierarchy —stifle the very benefits diversity could provide. As firms grow, they fragment into specialized departments and geographic units. Silos harden, communication turns formal and filtered, and office politics or turf wars take hold. In this environment, a diverse executive team often struggles to share unique insights and reach decisions. Consequently, larger companies spawn features that intensify social categorization, as leaders may identify more with their division than with the enterprise overall, breeding rivalries and silo thinking instead of cross-functional synergy. Leadership styles also shift toward directive, top-down approaches; with tens of thousands of employees, participatory dialogue — the kind that lets diverse ideas surface and integrate — gives way to command-and-control methods that limit open exchange. Scale itself creates communication barriers as information filters through layers of management and distant divisions, raising the odds messages will be lost, mistranslated, or unshared. Turf defense may stifle any energy for collaboration. Finally, sheer complexity exacts a toll: Fortune 100 companies juggle countless products, markets, and stakeholder demands. Competing sub-organizational goals erode identifying shared objectives and organizational loyalty. In large, complex companies, the coordination burden drowns out any productive creative friction across diverse executives, making unified collaboration difficult and erasing expected performance gains. Of note, recent research reported that 78 percent of senior leaders consider breakdowns in cross-unit collaboration a major problem, yet few feel effective at solving it4. Legacy structures in large firms make dismantling silos difficult, despite enthusiasm for cross-functional collaboration. Thus, large firms frequently fail to capture the diversity gains of smaller ones: diversity on paper, but not effective in practice. Turning Friction into Fuel: How Long-Tenured Teams Unlock Diversity’s Value Are big companies doomed to miss out on the upside of diverse teams? Not at all. Some large organizations do manage to consistently harness cross-functional expertise — the key differentiator is integration. Our research on top management teams show that firms succeeded in leveraging functional diversity when management teams had levels of shared experience. In our study of 4,500 organizations, when a top management team had long tenure — at least 7 years of time on their team — the negative effects that diversity had on large firms disappeared. A well-integrated top management team is not about everyone thinking alike; it’s about having strong trust, communication, and mutual understanding despite their different backgrounds. Building that kind of cohesion takes time and leadership. Executives who have been through battles together can learn how each other thinks and “bridge semantic gaps” between their functional languages. Studies show shared experience provides time for interpersonal trust and psychological safety to develop, enhancing information exchange and integrating diverse knowledge. This is the secret sauce turning diversity into performance gold: team members feel safe to speak up and have the trust to truly listen to one another, making differences a source of strength rather than division. Case in Point: Ford’s “One Team” Revolution Ford’s dramatic turnaround under CEO Alan Mulally (2006-2014) shows how to turn diverse leadership into real advantage5. The company was a siloed disaster — “warring factions” nearly bankrupted it. Mulally assembled a functionally diverse executive team but didn’t stop there: he created weekly “Business Plan Review” meetings where leaders openly shared progress and issues using color-coded charts. Initially, everyone reported only green (“all good”). Mulally publicly praised the first honest red report and rallied the team to help solve it, sparking psychological safety. Silos eroded as executives began collaborating across functions. The payoff: Ford avoided the 2008 bailout, became the only profitable Detroit automaker during the recession, and reversed massive losses. This cultural shift “from toxically competitive to collaborative”6 earned big dividends for the company, and showing even in a huge organization, leaders can foster integration so that diversity delivers on its promise. Our research finding about long-tenured teams aligns with stories like Ford’s. If your top management team hasn’t had time to gel, all the diversity in the world might not help — it could even hinder, as members struggle to understand each other. Given time (or deliberate team-building efforts), diverse teams become far greater than the sum of their parts. Critically, stability in a leadership team can amplify the value of diversity, undercutting the advice to reshuffle executives or bring in “fresh blood” frequently. Fresh perspectives are valuable, but don’t underestimate the power of a team that has learned to play well together. Revisiting the Business Case: Correlation, Causation, and the New Evidence It’s worth considering what the broader evidence tells us about diversity’s impact on performance, beyond individual stories. In the mid-2010s, influential consulting reports — especially McKinsey’s “Why Diversity Matters” series — suggested companies with greater gender/ethnic diversity in leadership tended toward stronger financial results. The 2015 report noted these findings were correlation (in smaller print) which does not imply causation, while still highlighting a potential link to success7. One limitation is that the studies measured diversity at the end of a period and linked it to earlier financial performance — which could mean successful companies were simply better positioned to attract and promote diverse leaders, rather than diversity directly driving the gains. Recent rigorous studies have poured cold water on the simplistic “diversity = higher profit” narrative when it comes to demographic diversity in large firms. For example, a 2024 academic study of all S&P 500 companies (McKinsey’s focus) found no clear link between executive team racial/gender diversity and future financial performance. They concluded that the oft-touted business case for demographic diversity is overstated, and they questioned McKinsey’s results, suggesting that McKinsey likely got the direction of causality wrong8 . In short, earlier claims that simply diversifying a leadership team will automatically boost your bottom line are not backed by solid evidence. It’s more complicated. Our research differs from McKinsey’s simple study (and its subsequent debunking) by examining functional diversity across a large sample and long time period. We tracked performance changes as team compositions evolved and applied methods to address reverse causality and confounders. This allowed us to isolate functional diversity’s true effects: it can improve firm performance, but context is key. In smaller firms, it provides a direct boost; in larger ones, it requires conditions like team integration and tenure to yield benefits. This nuanced view affirms diversity’s value while steering clear of one-size-fits-all claims. Executives need to pursue diversity with both optimism and realism. Assembling varied experts or demographics alone guarantees nothing — results depend on how the team is managed and the organizational context in which it operates. Making Diversity Work: Leadership Lessons for Harnessing Differences How can leaders of organizations — big or small — leverage the power of functional diversity while avoiding its pitfalls? A few actionable lessons emerge: Foster a “One Team” Culture: Follow Mulally’s Ford example—break silos with regular forums where leaders explain issues in plain language. Rotate chairs or use facilitators to ensure no single function dominates. The goal is to instill an ethos that we win or lose together, not in isolation. When every executive feels responsible for collective problems, not just their silo, diverse thinking converges into unified action. Invest in Integration Mechanisms: Don’t leave cohesion to chance. Use cross-functional projects, offsites, co-location, and mixed org structures (e.g., embedding analysts across teams) to create shared experiences and daily exchange. These build trust and mutual understanding, mimicking small-firm closeness for better collaboration. Practices like mixing functions within physical spaces can simulate the close-knit feel of a smaller firm, encouraging daily knowledge exchange across specialties. As research suggests, familiarity breeds collaboration in this context – when people know each other well, they communicate more freely and productively. Mind Your Team’s Tenure Balance and consider how the CEO fits in: In large companies, frequent reshuffles can disrupt cohesion — stability helps diverse teams gel and develop. When adding new blood, use mentoring or overlap periods. Also, pay attention to leadership development: consider cultivating CEOs who have broad functional experiences in their career (so-called “generalist” CEOs). We found a CEO with a broad background can somewhat substitute for team diversity — perhaps due to a bridging ability to speak everyone’s language. When you don’t want (or can’t form) a heterogeneous team, a cross-functional polymath at the helm can be an alternative way to have multiple perspectives at the top. Measure and Adapt: Finally, treat the impact of diversity as a hypothesis to continually test and refine in your own company. Maybe you’ve improved gender or functional diversity in your leadership team — track how it correlates with outcomes over time, and gather feedback on team dynamics. If results aren’t what you hoped, dig into why: Do people feel included or is the diverse team just “for show”? Are there communication clogs you can clear? By treating this as a learning journey, you avoid the extremes of blind faith or cynical dismissal. Instead, you’ll incrementally discover what mix of talents and dynamics truly drives performance in your context. Conclusion: Beyond Buzzwords to Better Performance “Diversity” should not just be a box to check or a slogan to trot out at shareholder meetings. It’s a capability — the capability of an organization to think differently within itself, to have constructive debate, and to approach challenges from multiple angles. As we’ve seen, that capability can create tremendous value, but it flourishes under the right conditions: a culture of integration, a size that allows voices to be heard, and leadership that actively cultivates cohesion and trust. For a small enterprise, the lesson is clear: embrace functional diversity early. Your small size is an advantage — you can meld your all-stars into a tight-knit, cross-disciplinary unit that outthinks and outmaneuvers bigger rivals. For a large corporation, the task is more delicate: don’t assume that diversity automatically yields dividends. Be intentional in breaking silos and forging a one-team mentality in your upper echelons. It may take time and persistence (old habits die hard, as Ford’s story shows), but the payoff is a leadership team that can actually capitalize on the wealth of knowledge it possesses. In the end, the debate about whether diversity matters for performance is settling into a more mature phase. It’s not if it matters — it’s when and how it matters. The newest evidence suggests that different backgrounds do matter — they can be the catalyst for superior performance, but only when combined with unity of purpose and effort. For businesses of all sizes, the mandate is not just to have diversity, but to enable it. The companies that figure this out will enjoy more innovative strategies, more robust decisions, and yes, likely better financial results over the long haul. Those that don’t will continue to wonder why “diversity programs” didn’t magically make a difference. In the words of an old proverb, “If you want to go fast, go alone. If you want to go far, go together.” We might add a modern corollary: If you want to go further, go together with people who aren’t just like you — and take the time to truly come together. That’s the savvy way to leverage diversity for performance, turning what could be friction into the engine of future success. References Frances Fabian et al., “When Does Top Management Team Diversity Matter in Large Organizations?” Journal of Organizational Behavior, accepted August 24, 2005. Due to the fact we study public firms, small firms in our sample still have at least a few million dollars worth of assets. Roni Reiter-Palmon et al., “Teams in small organizations: Conceptual, methodological, and practical considerations,” Frontiers in Psychology 12 (2021): 530291. Sharon Ceurvorst et al., “Why Cross-Functional Collaboration Stalls, and How to Fix It.” Harvard Business Review, June 24, 2024. Ernest Gundling, “Disruption in Detroit: Ford, Silicon Valley, and Beyond,” California Management Review Case, January 1, 2018. Tom Relihan, “Fixing a toxic work culture: Breaking down barriers,” MIT Sloan Management Review, May 29, 2019. Vivian Hunt et al., “Diversity Matters,” McKinsey&Company, February 2, 2015. Jeremiah Green and John Hand, “McKinsey’s Diversity Matters/Delivers/Wins Results Revisited,” Econ Journal Watch 21, no. 1 (2024): 5–34.

This article is adapted from Making Organizational Culture Great: Moving Beyond Popular Beliefs published by Columbia Business School Publishing (c) 2026 Jennifer A. Chatman and Glenn R. Carroll. Used by arrangement with the Publisher. All rights reserved. Popular Beliefs About Culture Culture baffles even the most experienced managers. Even those who take on the challenge of leveraging their organization’s culture for strategic success often feel mystified, uneasy, or skeptical. There are plenty of good reasons for this discomfort. First, culture is not a tangible phenomenon that you can readily see. Second, managers typically receive little or no training in creating or managing culture, unlike their training on tasks like manipulating a spreadsheet or reporting financial outcomes. And few managers are social scientists. Yet the stakes—using culture to accelerate your organization’s success or, conversely, letting cultural inertia doom your organization—are high. Most people—especially executives and other top leaders—believe that culture matters enormously for how an organization operates and performs, both in the short and long term. Consider the findings of prominent management consulting firms. For example, a 2021 survey of 3,243 executives in forty-two countries by consulting firm PwC found that “81 percent of respondents who strongly believe their organization was able to adapt during the 12 months before our survey was conducted also say their culture has been a source of competitive advantage.” Similarly, Deloitte’s survey of 1,308 adults and executives in 2012 found that “94 percent of executives and 88 percent of employees believe a distinct workplace culture is important to business success.” A study by Korn Ferry found that “91 percent of executives agree that improving corporate culture would increase their organization’s value” and “80 percent of executives ranked culture among the five most important factors driving valuation.” Likewise, Heidrick & Struggles’ 2021 survey of 500 CEOs at companies with a minimum of $2.5 billion in annual revenue found that “82 percent of CEOs . . . surveyed said they had focused on culture as a key priority over the past three years.” Academic researchers report similar findings. John Graham and colleagues surveyed 1,348 CFOs and other finance executives around 2020. They found that “91 percent of executives consider corporate culture to be ‘important’ or ‘very important’ at their firm.” Similarly, Glenn Carroll and Lara Yang surveyed 1,926 managers and nonmanagers in the United States about cultural beliefs, perceptions, and experiences. They found that about half the respondents reacted positively to the statement, “In general, culture is more important to organizational performance than strategy or operating model.” Despite the professed importance of culture, a Gallup poll found that only 21 percent of employees report feeling connected to their company’s culture. Our Approach We wrote this book to help managers develop and manage culture so they can improve their organization’s performance. We do so by helping to sort out what’s what with respect to culture, to consider several of the most salient popular beliefs about culture, and to offer our evaluations as professional social scientists, one of us (Chatman) a psychologist and the other (Carroll) a sociologist. We have been researching organizational culture for decades. Our main goal in writing this book is to offer guidance on how to manage organizational culture effectively to those who are responsible for leading and directing organizations and the teams of people within them, including divisions, departments, and other units. We recognize at the outset the difficulty of try- ing to define culture. Academic definitions often extend broadly to include symbols, behaviors, norms, values, and language. For example, pioneering culture researcher Ed Schein defined organizational culture as “the pattern of basic assumptions which a given group has invented, discovered or developed in learning to cope with its problems of external adaptation and internal integration, which have worked well enough to be considered valid, and therefore to be taught to new members as the correct way to perceive, think and feel in relation to those problems. It is the assumptions which lie behind values and which determine the behavior patterns and the visible artifacts such as architecture, office layout, dress codes, and so on.” Yet we view it as counter-productive for managers to worry about definitional debates. We suggest using a simpler, more straightforward definition of culture as “a system of shared values that define what is important, and norms that define appropriate attitudes and behaviors for organizational members.” Culture can also be hard to identify “in the wild.” To see culture in action can be like trying to spot camouflaged animals in the jungle. Adding to the challenge, culturally relevant behaviors can be ambiguous, frequently spawning multiple interpretations. And members of an organization’s culture can claim to have a certain culture, but the reality of that culture can be quite different from what people say it is. Mainly, we aim to offer guidance on how to manage organizational culture—ranging from crafting a culture that helps an organization execute on its strategy to ensuring that the culture adapts over time. We do so in a particular way—by sorting out what is true about culture, what’s not true, and what appears ambiguous or unresolved.9 We believe that this approach will enable managers to prioritize what really matters, to understand what is consequential, and to know what to ignore and leave behind. To illustrate our approach, consider the issue of measuring culture quantitatively. Many managers wonder whether measuring culture is a good idea, and if so, how and when they would do so. Others wonder whether their hiring processes should evaluate a person’s fit to the culture, and if so, how they can avoid bias and discrimination in the process. And, of course, managers wonder about culture’s impact on organizational performance and how they can ensure that culture helps rather than hinders people trying to accomplish organizational goals. These questions often challenge managers as well as social scientists. But they become even harder to answer, if not impossible, without a solid understanding of the behavioral realities of culture, which requires looking beyond the popular beliefs to find what’s true about culture. Strong Culture Organizations We pay particular attention to a culture’s strength, a widely used social science term that is sometimes misunderstood. Specifically, social scientists define a strong culture organization by two pronounced features. First, its members hold a high consensus around the appropriate norms, values, and beliefs of the organization. In other words, people agree about “the right thing to do at this organization.” Second, members display a high intensity of commitment to those norms, values, and beliefs, such that people will act on their own to ensure that others comply. Imagine being taught on the assembly line the “right” way to do things by your fellow worker or being scolded by your peer when violating a normative expectation of timely attendance at meetings. In both cases, the targeted employee is being instructed and sanctioned by a peer rather than a boss. This self-managed aspect of strong culture organizations is part of their appeal, and systematic research (reviewed in chapter 7) shows that strong culture organizations indeed require fewer managers to operate effectively—operationally, they are simply more efficient. Note that by this definition, a strong culture organization does not depend on any specific norms or practices (often called “cultural content”) to make it strong—all that’s required is high agreement and high intensity. Another way of saying this is that cultural strength is independent of cultural content. Accordingly, you can find examples of strong culture organizations with virtually any cultural content. Indeed, in this book we will review examples of strong culture organizations engaged in manufacturing, service delivery, research, terrorism, religion (including cults), policing, military activities, and more. We will see strong culture organizations that are large and small, old and newly founded, across a variety of industries and operating in many different countries. For example, among commonly recognized strong culture organizations are the following well-known organizations: Southwest Airlines, with its breezy culture of fun and teamwork The Unification Church, aka the “Moonies,” a religious cult that draws people in and won’t let them go easily Nordstrom, the Seattle-based department store known for exceptional customer service Goldman Sachs, the long-successful investment banking firm that drives performance through information sharing Navy SEALs, Green Berets, Special Weapons and Tactics (SWAT) teams, military-like special forces, who perform highly specialized and immensely difficult tasks for national defense Amazon, the internet-based retailer and cloud service com- pany who seeks to provide consumers with anything they might want to buy online Netflix, the video-streaming service whose culture has enabled it to transform its business model from sending DVDs in the mail to now producing its own entertainment content Google, the information technology company built on inter- net search who selects people based on sheer curiosity Uber, the ride-hailing platform that dominates most US urban markets Five Popular Beliefs We organized this book around five common popular beliefs about culture. We focus on these specific beliefs because they come up most frequently in our consulting activities and classroom teaching. The beliefs capture many of the key challenges that managers face in using culture to improve organizational performance and to sustain performance at high levels. The discomfort that many managers feel about culture often leads to misdirected efforts to learn more about culture or to engage consultants and other experts. While we applaud any attempt to gain more knowledge, we think that the biggest challenge that leaders face in managing culture is not intellectual but behavioral. Once you learn to ignore the munificent and often imprecise babble about culture—the underlying forces involved in building and maintaining a strong culture hold little mystery, at least to social scientists. Cultures get built and sustained through a variety of well-studied and well-known social and psychological processes. For the most part, social scientists do not question or debate the ways that these processes operate; they reside in the scientific canon as accepted facts. Most important, these processes are not mysterious or technical: You can readily learn and remember them. For example, the relevant managerial levers include culturally selective hiring, intense early socialization, aligning compensation and other incentives, and communicating expectations throughout the organization. The payoff can be substantial because under- standing and implementing these processes will enable you to enhance your organization’s well-being and performance. By contrast, what is difficult—exceptionally so in our estimation—is the ability to act and behave consistently in ways that advance your goals as a manager in using these processes. Acting consistently day in and day out, meeting after meeting, activity after activity, in the presence of many different people holding many different positions and playing many different roles, requires self-discipline, deliberateness, and personal presence. Jack Welch, the highly successful long-term CEO at GE, famously said that good leaders had to be “relentlessly boring.” Welch was not advocating that a leader be boring when speaking or acting (he passionately believed the opposite) but he appreciated that if an intelligent and engaged leader repeats the same message consistently hundreds, perhaps thousands, of times, then it will likely get boring to the leader himself. Welch was warning against being harmfully inventive by modifying the message to make it interesting to the leader. A second difficulty lies in ensuring a comprehensive approach to managing culture. That is, leading through culture involves using a variety of managerial levers that affect a variety of processes. While there may be some absolute no-no’s, there is no magic bullet, no single way to build and sustain an organization’s culture. You must attend to several or many levers and processes at once if you want to manage the culture effectively. It’s not just about incentives, training, or culturally selective hiring—it’s about orchestrating a wide range of the levers at your command. The challenge rests with juggling many balls to the same end, some of which may be easy for you and some which you will find hard. Finally, offering a coherent narrative about these consistent and comprehensive practices ensures that members of your organization understand without ambiguity why you wish to cultivate a particular culture, with specific behavioral norms. What do various groups in the organization—the executive team, managers, individual contributors, and others—think you and they will gain by following this particular culture with these values and norms? The narrative contains both formal scripted chapters as well as informal spontaneous ones. Cultural coherence provides the logic for coordination across organizations, something essential for getting big things done. So, in our view, success in managing culture does not require you to become a rocket scientist—defining and figuring out difficult unsolved problems. Instead, the challenge involves performing on point. Perhaps an orthopedic surgeon represents a better metaphor—a knowledgeable professional who executes time and time again in a consistent, comprehensive, and coherent way. Managers need to behave consistently, comprehensively, and coherently in scientifically known ways to get the results that they hope for in managing organizational culture. Our goal in writing this book is to demystify culture, to offer clarity about the known and proven ways of leading and managing an effective culture. The book will demonstrate that managing through culture typically differs from conventional management in numerous ways. For example, leading through culture involves culturally selective hiring for fit rather than just focusing on skills. New hires are socialized to the culture, and motivational messages aim to inspire instead of offering higher pay. Leaders are often treated like peers, information is shared widely, and peers are involved in supervision as much as bosses. Rules in strong culture organizations also tend to general and generative rather than specific and detailed. Strong culture organizations typically manage through social control—peer pressure or normative sanctioning—rather than heavy doses of formal control, consisting of rules, policies, and defined procedures. Editorial Note This post is adapted from a chapter of the book "Making Organizational Culture Great: Moving Beyond Popular Beliefs" by Jennifer A. Chatman and Glenn R. Carroll (c) April 2026 Columbia Business School Publishing. Used by arrangement with the Publisher. All rights reserved.

Managers deciding where and how to automate supply chains typically anchor their analysis on internal metrics: labour costs, throughput targets, return on investment. Yet in many emerging economies, a parallel transformation is unfolding that makes those static calculations obsolete. Governments are not simply fixing potholes or adding ports; they are fundamentally rewiring logistics infrastructure, integrating previously disconnected modes, and exposing real-time data through digital platforms. India’s PM Gati Shakti a GIS-based coordination system linking 16 ministries to plan railways, roads, ports, inland waterways and logistics parks as one multimodal network illustrates this shift at scale. Brazil’s PAC infrastructure programme and Indonesia’s logistics modernization efforts follow similar logics. The managerial question is no longer “should we automate?” but “how do we design automation strategy when the external logistics system is not fixed but changing under our feet?” This article offers a framework for aligning firm-level automation with state-led infrastructure transformation. Drawing on India’s logistics automation market projected to grow from USD 1.88 billion in 2024 to over USD 8 billion by 2033, alongside an estimated 80% of warehouses adopting some automation by 2030 the framework identifies when automation amplifies infrastructure gains and when it becomes stranded investment. The core insight: automation returns depend less on “how much technology” and more on timing, location and complementarity with external policy execution. Managers who treat automation roadmaps as independent of infrastructure maps risk deploying expensive assets in precisely the wrong places at the wrong moments. The Puzzle: Why Similar Automation Investments Pay Off Differently Consider two warehouses in India, each investing roughly USD 2 million in semi-automated sorting, put-to-light systems and warehouse management software. The first sits in an industrial estate 60 kilometres from the nearest rail link, relying on road freight through congested corridors. Power is unstable; broadband patchy. The operator cannot access real-time data on vessel berthing, train movements or port congestion because the facility predates government digital platforms. When customer contracts shift or volumes drop, the sorter becomes a fixed-cost burden the firm cannot easily redeploy. The second warehouse is located inside a Multi-Modal Logistics Park (MMLP) co-designed with rail sidings, highway access and dedicated power. The operator plugs warehouse management and transportation systems directly into India’s Unified Logistics Interface Platform (ULIP), which exposes over 1,800 data fields from 41 government systems via APIs vessel schedules, rail rake visibility, customs documentation, e-way bills. When disruptions hit, the firm can reroute shipments across rail, road or coastal modes because the infrastructure and data to do so exist in real time. The automation investment here amplifies gains from better connectivity, lower dwell times and modal flexibility. Both firms “automated.” Only one captured the complementary value from infrastructure transformation. This is not an India-specific problem; it is a structural challenge wherever states are rewiring logistics at the same time firms are automating operations. Framework: The Automation–Infrastructure Alignment Matrix To navigate this environment, managers need a simple but robust heuristic. The Automation–Infrastructure Alignment Matrix maps automation intensity against logistics infrastructure quality and policy alignment. Figure 1: Automation–Infrastructure Alignment Matrix. The vertical axis represents automation intensity (low to high); the horizontal axis reflects infrastructure and policy alignment (low to high). The four quadrants capture distinct risk–return profiles and strategic choices for managers navigating infrastructure transformation. Quadrant 1: Stranded Automation (High Automation, Low Alignment) Firms here have deployed sophisticated automation in locations poorly served by infrastructure or excluded from policy coordination. Hardware spend exceeds 50% of total automation investment in India’s logistics market, and much of it sits in precisely this quadrant. A 3PL operating a highly automated warehouse off-corridor faces long, variable lead times that automation cannot compress because external bottlenecks dominate; limited modal choice, forcing reliance on congested road freight; and no access to real-time government logistics data, so planning systems operate with stale information. The economic risk is acute in volatile markets. When customer contracts shift common in India’s fragmented 3PL sector where multi-year contracts are rare firms cannot easily redeploy fixed automation assets. What looked like “strategic” investment becomes a sunk cost eating margin. Managerial implication: Avoid front-loading automation in locations where infrastructure quality lags. Treat such investments as options, not commitments pilot with modular, subscription-priced solutions until infrastructure clarity improves. Quadrant 2: Latent Potential (Low Automation, High Alignment) Facilities here sit in well-connected locations Gati Shakti corridors, designated MMLPs, export-oriented industrial clusters but have not yet automated meaningfully. This is the highest-return space for new automation investment because external enablers already exist. India has roughly 35% of logistics automation spend concentrated in Western India’s logistics hubs, but many smaller operators in those same hubs remain manual. They benefit from better roads, multimodal access and faster customs clearance, but they leave productivity gains on the table by not mechanising internal flows or digitising planning. Managerial implication: Prioritise these locations for rapid automation scale-up. IRR calculations anchored only on internal labour and error reduction understate true returns, because improved corridor speed, lower dwell times and modal flexibility compound automation gains. Quadrant 3: Cautious Optimisation (Medium–Medium) Most mid-sized firms cluster here: incremental automation in moderately connected locations. Operators adopt warehouse management systems, basic mechanisation and some analytics, but avoid big robotics bets. This is rational risk management in uncertain environments, but it also means firms are not positioned to exploit infrastructure breakthroughs when they arrive. Managerial implication: Build node-specific automation roadmaps tied to infrastructure timelines. When a corridor upgrade, MMLP commissioning or port expansion is confirmed, pre-position modular automation capacity to scale quickly once external bottlenecks clear. Quadrant 4: Policy-Leveraged Automation (High Automation, High Alignment) This is the strategic target zone. Firms here combine high automation intensity with strong infrastructure and policy alignment. They operate in or near MMLPs, plug into government digital platforms like ULIP and the Logistics Data Bank (which tracks 100% of India’s containerised export-import cargo via RFID), and co-invest in workforce skilling aligned with government training modules. Automation here acts as a force multiplier for public infrastructure, not a substitute. India’s e-commerce and export-focused FMCG sectors increasingly occupy this quadrant. With logistics costs estimated to have dropped from 13–14% of GDP historically to a 7.8–8.9% band recently driven by better infrastructure coordination automation in well-connected nodes delivers compounding returns: faster internal flows meet faster external corridors, and digital integration reduces planning blind spots. Managerial implication: Anchor major automation capex to this quadrant. Design investments as complements to policy execution, not independent bets. Sequence automation to follow infrastructure completion, not precede it. See Figure 2 for managerial implications. Table 1: Risk–Return Profiles Across the Four Quadrants Figure 2: Automation–Infrastructure Alignment Matrix Managerial Implication Three Complementarities That Determine Automation ROI Beneath the matrix sits a deeper structural logic. Automation returns depend on complementarity with three external assets the firm does not control: Network Complementarity: Physical connectivity quality multimodal links, corridor speeds, terminal throughput. PM Gati Shakti’s core promise is to move India’s logistics infrastructure from fragmented, single-mode planning to integrated, multimodal design. Firms automating yard management, gate systems or control towers inside MMLPs capture gains that isolated warehouses cannot, because trucks, trains and ships actually move faster and more reliably through those nodes. Data Complementarity: Access to real-time, system-level logistics data. India’s ULIP connects 41 government systems and exposes vessel berthing, rail schedules, port congestion and customs workflows via APIs. When a firm’s WMS or TMS integrates with ULIP, automated planning engines operate on current, accurate data rather than guesswork. Firms off the ULIP grid automate with one hand tied behind their backs. Human Complementarity: Availability of supervisors, technicians and planners who can interpret automated systems and manage exceptions. India’s National Logistics Policy explicitly targets workforce skilling through platforms like iGOT and logistics training in higher education. Warehouses in Gati Shakti-linked districts that co-invest in training retain talent and exploit automation more fully. Firms that automate without skilling face high attrition, manual overrides and brittle operations when disruptions hit. Managers should audit automation investments against these three dimensions. A robotics project scoring high on network and data complementarity but low on human complementarity will underperform; so will one that ticks the human box but sits in a poorly connected location with no ULIP access. Propositions From the framework and complementarities, three testable propositions emerge: Proposition 1: The return on warehouse automation investment is significantly higher in locations with high infrastructure and policy alignment (proximity to multimodal hubs, access to government digital platforms) than in otherwise similar locations with low alignment. Proposition 2: Automation projects that exhibit strong complementarity across network, data and human dimensions achieve greater operational resilience and lower stranded-asset risk than projects that score high on only one or two dimensions. Proposition 3: In policy-active environments, firms that sequence automation to follow infrastructure completion (option-based strategy) outperform firms that front-load automation commitments in advance of infrastructure clarity. These propositions offer clear hypotheses for future empirical work and immediate guidance for managers evaluating automation portfolios. Managerial Playbook: Four Non-Obvious Moves Move 1: Map policy execution timelines before finalising automation roadmaps Obtain corridor completion schedules, MMLP commissioning dates and digital platform rollout plans. Sequence automation to follow infrastructure, not lead it. In India, firms can access PM Gati Shakti’s spatial data layers through the national portal; similar platforms exist or are emerging in other infrastructure-active economies. Move 2: Treat automation as real options on policy delivery In high-uncertainty environments, deploy modular, subscription-priced automation (cloud WMS, robotics-as-a-service) that can scale quickly when infrastructure clarity improves. Avoid large, fixed robotics investments in locations where policy execution risk is high. Move 3: Build a “policy radar” function in supply chain teams Designate staff to track infrastructure announcements, budget allocations and digital platform rollouts. Front-load pilot automation in locations where the state is over-investing. In India, Western India accounted for over 35% of logistics automation spend precisely because Gati Shakti and earlier programmes concentrated multimodal investments there. Move 4: Co-invest in complementary workforce development Do not automate in isolation. Partner with government skilling programmes, vocational institutes and logistics academies to ensure supervisors and technicians can exploit automation. Firms that upgrade roles from manual pickers to robot operators, from paper-based planners to control-tower coordinators retain talent and sustain automation gains. Implications Beyond India The framework generalises. Any context where states are simultaneously upgrading hard infrastructure, integrating modes and exposing digital logistics data creates the conditions for policy-leveraged automation. Brazil’s logistics investment corridors, Indonesia’s logistics reform agenda and parts of Southeast Asia’s ASEAN connectivity push all fit this pattern. The managerial challenge is identical: how to time, locate and design automation so it amplifies rather than ignores or contradicts what the state is building. The alternative treating automation as a purely internal, firm-level decision produces the stranded investments, brittle systems and disappointed returns that characterise Quadrant 1. Managers who ignore the policy map deploy robots in the wrong places, at the wrong times, for the wrong reasons. Those who align automation roadmaps with infrastructure transformation capture compounding gains that static ROI models cannot see. India’s logistics automation market, racing from USD 1.88 billion to over USD 8 billion this decade, offers a live laboratory for this dynamic. The lesson is not “automate more” or “automate less.” It is: automate deliberately, with the grain of policy, in locations and at moments where external complementarities are strongest. That is how managers turn automation from a cost centre into a strategic lever and how they avoid the expensive mistakes visible across India’s warehouses, 3PLs and manufacturing clusters today.