The Blog TL;DR

Stakeholder-centered Design

Have you ever been surprised by the announcement of a new organization policy? Have you ever thought, “What were they thinking?” 

Effective organization and management design depends on an in-depth understanding of the key stakeholders of the particular organization, system, process, or project being designed. In many cases, taking a little more time to consult key stakeholders as part of the design process could prevent or at least mitigate some of these surprises and missteps.

Here at the Studio, we view organizations as systems of reciprocal value exchanges among stakeholders. The organization designer's task is to understand who the relevant stakeholders are, what they require, and then design activities and decisions that create value across the stakeholder system while minimizing avoidable trade-offs.

Six Stakeholder Groups

The Studio Stakeholder Framework includes six key stakeholder groups: customers, workforce, investors, suppliers and partners, society and communities, and the natural environment.

The first three form the core triad of any business: investors, workforce, and customers. Investors provide the resources to hire the workforce, build infrastructure, and develop products and services that solve customers' problems. If the firm does this well in customers' eyes, customers return, spend more, and tell their friends, creating a virtuous cycle that grows the firm.

A quick thought experiment illustrates the interdependence. If customers do not get enough value, they leave. If workforce members are dissatisfied, innovation, quality, and service can suffer. If investors do not receive a return commensurate with the risks, they look for other options and the cost of capital increases.

Add suppliers and partners, and you create an even more powerful system of value creation. Society and communities are inextricably linked to organizations through employment, products and services, taxes, regulation, and the larger economic system. The natural environment also finds a voice and influence through the other stakeholder groups and through mechanisms such as public policy and regulation.

Systems Perspective

Stakeholders often have competing requirements. If we view these requirements through the lens of the current system, the only solution might appear to be compromise and trade-offs, in other words, taking something from one stakeholder to serve another. A systems approach to design offers another possibility.

Once you adopt a systems view, it becomes clear that a business logic of value exchanges exists between stakeholders. Many examples show that you can make even more money by serving multiple stakeholders, including the environment. (e.g., Anderson, 1998; 2019). The task is to move beyond individual stakeholder needs to a true system of service that aligns the stakeholders in a virtuous cycle of value creation.

Don't Accept the Trade-Off Too Quickly

Organization designers should not accept apparent trade-offs prematurely. Instead, they should first look for leverage points and designs that expand the value available to multiple groups.

Sometimes our imagination fails us, and we create designs that include genuine trade-offs. When this happens, design the decision criteria to guide conscious trade-offs until we can imagine a better way.  

Systems thinking is difficult because the actions and results in organizations are often separated by time and space. A strategic decision made today may not produce observable results for months or years. At the same time, we teach and manage business function by function and then wonder why organizations develop functional silos.

The design challenge is to understand the systems of stakeholders and search for designs that create value for multiple stakeholders without shifting avoidable costs onto others.

Read the full article: https://www.orgdesignstudio.com/blog/stakeholder-centered-design

References

Anderson, R. C. (1998). Mid-course correction: Toward a sustainable enterprise: The Interface Model. Chelsea Green.

Anderson, R. (with Lanier, J. A., & Hawken, P.). (2019). Mid-Course correction revisited: The story and legacy of a radical industrialist and his quest for authentic change. Chelsea Green Publishing.

The Bench

Recently, there has been a lot of talk about using AI to transform work. In a recent Fortune article, Stephen Messer argues that organizations are adding AI to existing structures and processes instead of redesigning them to leverage AI's capabilities. He characterized this as “decoration” rather than transformation (Messer, 2026). He proposed that AI can eliminate many unnecessary activities. This is a fair point, but it doesn't justify his proposed solutions. 

Hammer 2.0?

We've seen this movie before. In a 1990 Harvard Business Review article by the same name, Michael Hammer proposed "Don't Automate, Obliterate," and we know how poorly that worked out (Hammer, 1990). Some of us who lived through the re-engineering era came away with a deeper understanding of the less visible coordination, judgment, and social functions that coordination work performed. The re-engineering tools were not the problem. The problem was that the essential coordination work of managing groups of people was much harder to replace than first thought. Simplistic solutions and slogans are poor substitutes for understanding the system. 

What can we learn from the past to improve the probability of effectively integrating AI to enhance our work vs. ruining it?

First, don't underestimate the value of organizational activities. Managerial coordination work is not merely a "relay" of information. Beyond relaying information, it involves sense-making under ambiguity, exception handling, developing people, holding them accountable, and so on.

Second, meetings are often much more than theater. While rituals like sales forecast meetings often seem like simple information exchanges, they often do commitment work for the goals and objectives that make up the organizational strategy. 

Both errors assume humans are machines and that their work is a simple engineering problem, ignoring the human and social aspects that make an organization effective.

Coordination

The managerial work of coordination addresses multiple useful functions beyond information exchange. We use a variety of coordination activities and artifacts, including meetings, approvals, and reports, that often serve multiple purposes. At a minimum, purposes include: information transfer, commitment, legitimation, buffering, sensemaking, capability development, and control. While information processing is the work AI most directly substitutes for, the others do not automatically follow.

An unstated assumption embedded in Messer’s subtraction argument is that the only function of the managerial coordination activities is information transfer. This is why he sees only theater in a forecast meeting. The better design move isn't to delete the ritual, but to analyze the functions each mechanism performs. In other words, before removing any activity, identify which functions it serves, and then decide where each one should go.

Three human-centered design considerations:

Commitment: a number a system infers is not a number anyone has committed to.

Legitimation: decisions need to be acceptable, not just correct, and acceptability is partly procedural. 

Development: junior people learn judgment by doing the low-value coordination work you're about to automate.

These considerations help turn subtraction from a slogan into a discipline, and they produce different answers in different organizations, which is what you want. Two firms running the same analysis on the same ritual will legitimately keep and cut different things depending on where their trust is thin and where their bench is weak.

Determining the Proper Fit

Evidence from a McKinsey survey suggests that simply adding AI is not enough and that changing the work itself is needed to improve performance, but it does not tell us how a particular organization should redesign that work (Singla et al., 2025).

A key challenge in incorporating AI into work design is ensuring the design fits both the nature of the activity and the capabilities and limitations of human and artificial contributors.

Three key aspects to consider are the nature of the activities, the capabilities and limitations of the contributors, and the design of the combination. Activities fall on a continuum from predictable to ambiguous (Repenning et al., 2018). Contributors fall on a continuum from deterministic to probabilistic. The intersections between the activities and the contributors are the design decisions.

Activities

Begin by exploring the nature of the activities. Start with three key questions:

Is the work well-defined with predictable results, or is it more ambiguous with the solution emerging from the process?
Is the work repeatable, or is it custom for each situation?
Are the criteria for decisions known in advance, or do they require informed judgment?

Stable, well-understood work can support more specification, routinization, serial flow, and mechanistic structures (Repenning et al., 2018). By contrast, work characterized by uncertainty and ambiguity generally requires more interaction, judgment, adaptation, and collaboration (Repenning et al., 2018). Creative activities such as strategy development, problem solving, and product design are ambiguous work where the output is not known in advance. While we structure these activities to improve the performance of the entity doing the work, the amount and type of structure are different from predictable processes.

Contributors

What are the relative natures of the contributors? How predictable are the entities doing the work? Contributors can include humans, AIs, and other technologies. In the past, we didn't need to explicitly say that machine automation was repeatable; it was axiomatic. Once we developed and tested the software, the same inputs produced the same results. It was predictable. That is no longer the case.

AI can sometimes look like automated factory technology while behaving more like a studio contributor. While the activities might be predictable, if the contributor is probabilistic, then so is the output. Humans and AI vary for different reasons, but both introduce uncertainty into work-system performance. This is a new design challenge when incorporating technology.

The combination of the nature of the activities and the contributors performing those activities provides a basis for determining the amount and type of structure that best fits the situation.

Combination

What system design will get the most from the combination? How should the activities and contributors be combined to produce the best-performing system? One approach is to design the workflow around where uncertainty occurs and how the system responds.

Based on the nature of the activities and decisions, what contributions are human, AI, and conventional technologies best suited to make? What design provides the best combination of humans and AI? As you explore the options, consider the sequencing of activities, decision rights, how much structure is needed and what type, and what kinds of checks (triggers, feedback, escalations) are appropriate. 

Examining the nuances of the many possible combinations is far more useful than relying on a simple rule: AI is probabilistic, so humans need to check it. Simply adding human review (humans in the loop) can become the AI equivalent of relying on final inspection to ensure quality: expensive, late, and often a substitute for better process design.

Organizational systems often mix well-defined and ambiguous activities performed by various combinations of humans, AI, and conventional technologies. Effective organization design matches the amount and type of structure to each activity-contributor combination and provides mechanisms for shifting modes when the nature or state of the work changes. The ability to move effectively among those modes is what allows the system to achieve both efficiency and adaptability.

In conclusion, Messer's point about subtracting work is a good one, but it works only when it follows an understanding of the system, and it is destructive when it substitutes for one.

References

Hammer, M. (1990). Reengineering work: Don't automate, obliterate. Harvard Business Review, 68(4), 104–112. https://hbr.org/1990/07/reengineering-work-dont-automate-obliterate

Messer, S. (2026, August 15). AI isn't changing how companies work. It's changing what a company is. Fortune. https://fortune.com/2026/08/15/stephen-messer-artificial-commonsense-ai-changing-company/

Repenning, N. P., Kieffer, D., & Repenning, J. (2018). A new approach to designing work. MIT Sloan Management Review, 59(2), 29–38. https://sloanreview.mit.edu/article/a-new-approach-to-designing-work/

Singla, A., Sukharevsky, A., Yee, L., Chui, M., & Hall, B. (2025, March 12). The state of AI: How organizations are rewiring to capture value. McKinsey & Company. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai-how-organizations-are-rewiring-to-capture-value

The Evidence

Which functions of teamwork can AI perform, and which still require humans? What does the empirical evidence say? 

Generating possibilities

A recent study of 791 experienced P&G professionals working on real product-development challenges found that AI could handle some functions traditionally done by human teammates (Dell'Acqua, Ayoubi, et al., 2026). They studied individuals and two-person cross-functional teams, with and without AI. They found that individuals working with AI produced solutions roughly comparable in quality to two-person human teams without AI. In addition, AI helped the participants move beyond their functional areas of expertise. For example, commercial and R&D professionals produced more balanced solutions spanning both perspectives.

What are the implications for organization design teams? AI may provide some of the insights and inputs we traditionally bring in other specialists to provide. This could help expand the range of possibilities and could be particularly useful for complex cross-functional organization design challenges. While the findings do not establish that AI can replace team members, they do show that AI can perform some of the functions traditionally provided by teammates and give design teams access to additional perspectives.

In addition to supplying perspectives that might otherwise require another specialist, AI may expand the range of organizational design alternatives the team considers. Thus, AI may change both how the design team works and what it can design.

Evaluating possibilities

Generating possibilities is not the same as judging them! While the P&G study found that AI substantially improved the quality of the ideas generated, it did not find that it improved people's ability to select the best ideas. The two-person human teams working without AI were best at selecting their strongest idea.

The research findings are varied and nuanced. A meta-analysis of 106 experiments found that human-AI combinations did not outperform the better of humans or AI working alone (Vaccaro, Almaatouq, and Malone, 2024). However, results varied substantially by task: combinations performed more favorably on content-creation tasks and showed losses on decision tasks. 

The implication for organizational design is the question itself. 

The findings suggest that the question we should be asking is not whether it should be human or AI, but which parts of the work each should perform. 

For example, many organizational systems include several activities, from generating ideas to exploring and combining to evaluating and ultimately deciding. Which of these activities and sub-tasks are performed best by humans, AI, or human+AI?

Using the evidence

How can we put this evidence to good use? A key challenge is avoiding unjustified design rules based on these useful findings. While well-designed experiments can provide strong evidence about what happened in a particular setting, that does not mean those findings will work in other contexts. Consequently, organization designers need to use insights on what worked, what didn't, and under what conditions with caution, and apply their own understanding and judgment. This is especially true of the research on applying AI to organizational activities. 

In a recent experiment, 758 BCG consultants performed tasks with and without AI assistance (Dell'Acqua, McFowland, et al., 2026). For tasks that fell within the AI's capability frontier, consultants performed substantially better with AI than without it. However, on tasks that fell outside the AI's capability frontier, consultants using AI were 19 percentage points less likely to reach the correct conclusion than those working without AI.

What makes this even more challenging is that AI's capability frontier is jagged, meaning that tasks that appear similarly difficult to us may fall on opposite sides of what AI can reliably do. This is one of several reasons AI is another design variable for organization designers, not a universal substitute for people. The allocation of human and AI to tasks will vary with the task, context, and technology.

Conclusion: The opportunity for organization designers is not simply to build smaller teams; rather, it is to redesign the division of cognitive labor.

References

Dell’Acqua, F., Ayoubi, C., Lifshitz, H., Sadun, R., Mollick, E., Mollick, L., Han, Y., Goldman, J., Nair, H., Taub, S., & Lakhani, K. R. (2026). The cybernetic teammate: A field experiment on generative AI and teamwork. Organization Science, 37(4), 1217–1242. https://doi.org/10.1287/orsc.2025.20702

Dell’Acqua, F., McFowland, E., III, Mollick, E., Lifshitz, H., Kellogg, K. C., Rajendran, S., Krayer, L., Candelon, F., & Lakhani, K. R. (2026). Navigating the jagged technological frontier: Field experimental evidence of the effects of artificial intelligence on knowledge worker productivity and quality. Organization Science, 37(2), 403–423. https://doi.org/10.1287/orsc.2025.21838

Vaccaro, M., Almaatouq, A., & Malone, T. W. (2024). When combinations of humans and AI are useful: A systematic review and meta-analysis. Nature Human Behaviour, 8, 2293–2303. https://doi.org/10.1038/s41562-024-02024-1