Capability Boundary
What should the business own, control, partner for or automate?
The traditional make-or-buy question is too narrow for a world where strategy, data, execution and technology can sit in different places. This model starts with established strategy and economics, then extends them for modern capability design.
Established thinking first. Our extension second.
The base logic comes from competitive strategy, transaction-cost economics, strategic sourcing, innovation-coupling research and real-options thinking. The extension is separating ownership, control, execution and automation instead of treating a capability as one indivisible choice.
Think of one specific capability.
Examples: analytics, customer support, recruiting, paid-media optimization, content production, software development. You do not enter it anywhere. Keep it in mind and answer the five questions below.
Does this capability create competitive advantage or strategically important learning?
Porter/Core Competence lens: is this activity meaningfully tied to differentiation, cost advantage, customer understanding or a capability the company needs to get better at over time?
Can this work be governed effectively through a market relationship?
Williamson lens: can quality be specified, suppliers compared, performance monitored and providers switched without excessive cost or dependency?
Can execution be separated from learning?
Pisano/Shih lens: if someone else performs the work, can the company still retain the knowledge required to innovate, improve the process and make better adjacent decisions?
How uncertain is the long-term answer?
Real-options lens: if economics, technology or demand are still moving, should the company learn first before making an irreversible commitment?
How much of the execution is repeatable and codifiable?
AI/automation overlay: can meaningful portions of the work be standardized, monitored and handled by software without giving away the strategic control points?
Own the strategic core. Be flexible about who performs the work.
- Strategy and decision rights
- Critical data and measurement
- Learning loop and customer insight
- Specialist delivery
- Temporary capacity
- Implementation support
- Repeatable analysis
- Monitoring and preparation
- Standardized production work
- Define the strategy, data, measurement and decision rights the company needs to control.
- Separate access to specialist execution from ownership of the capability itself.
- Pilot automation in repeatable work before redesigning the broader operating model.
The boundary of the firm is becoming more granular.
The old make-or-buy decision assumed ownership and execution were tightly connected. Increasingly they can be separated. A company can own the customer relationship, data, measurement and decision logic; partner for specialist execution; buy the platform; and automate repeatable tasks.
Activities still need to create advantage. Markets and firms still have different governance costs. Learning can still be lost when execution is separated from innovation.
AI, automation and platforms lower some execution and coordination costs, making it easier to separate the strategic core from the labor required to perform every task.
The better question is no longer simply “make or buy?” It is “which layers must remain inside the boundary of the firm for advantage, learning and control to compound?”
Why I think the traditional boundary decision is changing now.
Stanford’s 2026 AI Index reports 88% organizational AI adoption, while agent use remains in the single digits across nearly all business functions.
Stanford HAI →NBER research by Agrawal, Gans and Goldfarb shows that AI adoption can depend on how interacting tasks are designed and coordinated across the organization.
NBER →OECD research finds many firms still treat AI as plug-and-play even though adoption can require broader changes across operations and departments.
OECD →This model is an original synthesis and extension of established strategy and economics. It is not presented as a validated scoring instrument. Its purpose is to make the logic of a capability-boundary decision visible enough to challenge.