Prerequisites to check before scoring
| Model | Required conditions | If conditions are missing |
|---|---|---|
| Functional | Cross-functional integration mechanisms and an end-to-end process owner. | If a process owner is missing, first design handoffs and final decision rights. |
| Divisional | Units with sufficient resources, capable business management and separate performance data. | If resources or data are insufficient, retain functional management and pilot unit-level accountability first. |
| Matrix | Decision-based authority, capacity plans and an escalation route for disagreements. | If absent, do not recommend immediate implementation. Build capabilities or run a limited pilot. |
| Team-based | Teams can own outcomes and have sufficient skills and prioritization authority. | Resolve dependencies and delegate authority; simply renaming departments as teams is insufficient. |
| Network | Measurable outputs, capable partners, continuity plans and a value-chain owner. | Retain critical work internally or pilot one manageable link in the chain. |
An evidence-based selection process
- Define the design level. Specify whether the scope is the company, business unit, function, product or partner chain. Establish the baseline and objectives within the same scope.
- Gather evidence. Objectives and customers; product portfolio; value streams; decision rights; workload and capacity; management capabilities; data; costs and practical constraints.
- Screen mandatory prerequisites. Do not recommend immediate implementation of an option that fails the prerequisite checks. Record what must change to qualify.
- Compare at least two feasible options. Options may combine a base structure with teams, shared services or a network within a defined scope. Include improving the current structure as an alternative.
- Score with evidence and test sensitivity. Adjust weights to reflect strategy and check whether rankings change. An overall score does not replace mandatory prerequisites.
- Pilot and measure before and after. Choose scope, ownership, budget and a suitable evaluation cycle. An 8–12 week pilot may be appropriate if the work cycle allows; this is an implementation suggestion, not a mandatory standard.
A proposed advisory scoring scale
Use a 0–4 scale: 0 = clearly conflicts with objectives or does not meet requirements; 1 = major changes required; 2 = usable after gaps are addressed; 3 = suitable with supporting evidence; 4 = strong fit supported by pilot results or operating data. If information is missing, record “insufficient data”; do not automatically assign zero or produce a recommendation score.
| Criterion | Suggested weight | Validation question |
|---|---|---|
| Fit with objectives and competitive approach | 25% | Which two or three strategic priorities does the model support, and what trade-offs does it require? |
| Fit with value streams and customers | 20% | Who owns the end-to-end outcome? How many handoffs, dependencies and waiting points remain? |
| Leadership capability and decision rights | 15% | Can managers and units competently exercise the authority delegated to them? |
| Total cost within budget | 10% | Is there sufficient funding and capacity for transition and operation? Have hidden costs been included? |
| Quality and risk controls | 10% | Are oversight, business continuity and data protection adequate for the work? |
| Speed and adaptability | 10% | How quickly can work priorities change while maintaining quality? |
| Culture and collaboration capability | 5% | Do behaviors, incentives and trust support the proposed structure? |
| Data readiness and AI collaboration | 5% | Are suitable data, accountable owners and AI output measures available? |
Fit score out of 100 = sum of (weight × criterion score / 4)
Weights total 100. This is a proposed decision aid, not a scientifically validated scale. Agree weights before scoring; record evidence and assessors for every score. Do not interpret fit scores as probabilities of success.
Budget comprehensively and measure performance
For the same evaluation period, total costs include transition, management staffing, tools and data, training, coordination, shared services, partners, AI, and estimable losses from poor quality or disruption. Compare options under consistent volume and quality assumptions; separate one-off and recurring costs and avoid double counting.
Use a balanced set of measures: processing and decision time; conforming output per resource; quality and rework; customer outcomes; management or coordination costs; and workforce capability and workload. Establish baselines, targets, data sources and accountable owners. Record changes in demand, staffing or technology when comparing before and after, so improvements are not attributed solely to the new chart.
AI collaboration across all five models
AI may be represented as a coordinating entity with a name, tasks, capabilities and service commitments, such as an “AI Coordination Unit.” Place it within the chosen base structure. Every task still requires a human business owner and an authorized approver. This is a design proposed by the author, not a claim that all 15 example companies use the same AI arrangement.
| Base structure | Proposed AI placement and collaboration |
|---|---|
| Functional | A shared AI coordination unit serves functions, each with a business contact. Apply data- and task-specific permissions; avoid making AI another approval layer. |
| Divisional | Shared AI platforms and policies; business units own use cases and results. Avoid duplicate tool purchases and incompatible data standards across units. |
| Matrix | AI consolidates plans, resource loads and conflict alerts for both dimensions. It does not independently resolve authority disputes; retain human ownership and decision logs. |
| Team-based | Each team uses task-specific AI agents connected to a shared AI coordination unit. Measure conforming outputs and turnaround; do not assess productivity solely by the number of AI-generated tasks. |
| Network | AI supports demand, schedule and quality coordination between partners. Share only authorized data, retain a value-chain owner and provide continuity arrangements for AI or partner disruption. |
AI entity information fields ID and name; service scope; tasks and outputs; human owner; operating unit; data sources and access rights; action permissions and escalation points; service commitments; costs; acceptance criteria; logs; fallback arrangements.
Scope and use of sources
This advisory comparison draws on five reference diagrams from the book How Business Works (pages 68–77), five foundational websites C1–C5, supplementary sources C6–C7, and 15 company examples E1–E15. Historical percentages in the book are not reused as current data. Model descriptions are synthesized; adoption conditions, solutions, measures and scoring methods are practical proposals that require validation in each business.
- These five models are reference types within this comparison, not an exhaustive catalog of organizational structures. Centralization and decentralization are design dimensions that can accompany any of the five.
- There is no universal headcount, budget or management-layer threshold across industries. Cost comparisons describe cost mechanisms, not quotations or standard rates.
- Each model is illustrated by one representative industry cluster and three companies. This is not a complete list of applicable industries. Businesses within the same industry can use different models.
- A indicates a direct description by the company or a leader at the date stated in the source. B indicates localized evidence from role descriptions or recruitment. These labels assess support for a claim, not company quality.
- Older sources provide dated examples, not claims that structures remain unchanged on 20 October 2026. Network examples support the value-chain model, not the entire internal reporting structure.
Sources and supporting evidence
Getting organizational redesign right
2015 · Accessed 20 October 2026
Design around strategy, diagnose the current state, define roles before people and measure outcomes.
Organization Design
Service page, accessed 2026 · Reviewed 20 October 2026
Start with performance gaps; design behaviors, accountabilities and implementation mechanisms together.
Who has the D
2006 · Accessed 20 October 2026
RAPID decision roles; reporting lines alone do not establish who decides.
Operating Model Transformation
Accessed in 2026 · Reviewed 20 October 2026
Operating models encompass work, processes, technology and the people or digital entities performing it.
Star Model
Accessed in 2026 · Reviewed 20 October 2026
Align strategy, structure, processes, rewards and people.
Designing Matrix Organizations that Actually Work
Book introduction, 2012 · Accessed 20 October 2026
Matrices require aligned coordination, planning, objectives and rewards.
Discover the Spotify model
Accessed in 2026 · Reviewed 20 October 2026
Autonomous teams and professional communities; avoid copying labels without the supporting culture.
How Apple Is Organized for Innovation
HBR 2020, copy provided by Apple · Accessed 20 October 2026
Functional organization and expert leadership at the time of the article.
The 21st century corporation a conversation with Brian Chesky of Airbnb
2021, describing the 2020 change · Accessed 20 October 2026
A direct CEO interview about Airbnb’s transition to a functional organization in 2020.
No Matrix Organization
Published handbook, accessed 2026 · Reviewed 20 October 2026
Self-described functional management with one formal manager.
Corporate Structure
Published structure page, accessed 2026 · Reviewed 20 October 2026
Category-based SBUs, performance accountability and market-support arrangements.
At a glance
Published page, accessed 2026 · Reviewed 20 October 2026
Four category-based business groups on the reviewed page, rather than the older five-group chart.
2025 Operating segments and information by products restated
23 March 2026, effective 1 January 2026 · Accessed 20 October 2026
Three geographic zones and two globally managed businesses; describes organization as well as segment figures.
Leadership How to create a Great Team
9 June 2023 · Accessed 20 October 2026
An Operations leader describes a matrix within their own area of responsibility.
Head of Lifecycle Digital and Software Smart Infrastructure Buildings
Recruitment posting indexed in 2025 · Accessed 20 October 2026
A role description refers to a global matrix; localized evidence from an indexed copy, not a group-wide chart.
Head of Marketing and Communications BCSS
Recruitment page, accessed 2026 · Reviewed 20 October 2026
A multi-geography matrix context at BCSS, evidenced within the scope of the role.
Amazon Two Pizza Teams
Published page, accessed 2026 · Reviewed 20 October 2026
Cross-functional teams own a service end to end; the source also discusses limits and duplication risks.
How should you structure your engineering team
28 August 2019 · Accessed 20 October 2026
Leadership describes teams combining design, engineering and product; a time-specific example.
Tech Migrations the Spotify Way
2020 · Accessed 20 October 2026
Autonomous engineering teams and coordinated shared changes; does not establish that the 2012 chart remains unchanged today.
Form 10 K fiscal 2026
Fiscal year ended 31 May 2026 · Accessed 20 October 2026
Most manufacturing is performed by independent contractors, supporting the value-creation network example.
Sourcing and Supply Chain Annual Report 2025
2025 report · Accessed 20 October 2026
Almost all production is outsourced to independent partners alongside logistics coordination.
Strategy business model and value chain Annual Report 2025
2025 report · Accessed 20 October 2026
The brand owner coordinates external manufacturers, suppliers and logistics.
Reference diagrams: Functional, Divisional, Matrix, Team-based and Network structures from the book How Business Works. The illustrations in the comparison are a new set of vector graphics created by Evera.