By George Alifragis and Karthiga Ratnam
The question that has governed enterprise technology strategy for 30 years just became obsolete.
For decades, CEOs and CIOs have wrestled with the same strategic fork in the road whenever new technology emerged: Should we build or should we buy? This question shaped ERP implementations in the 1990s, cloud migrations in the 2010s, and digital transformations that began in the 2010s and continue today. The assumption has always been the same—building enables differentiation, buying enables speed.
But AI breaks this dichotomy completely. With the rise of foundation models, agentic workflows, and domain-specific autonomy, the old question no longer makes structural sense. In this new age of AI, every company will both buy and build.
The strategic question has become: "Where do you buy leverage, and where do you build moats? And how do you orchestrate intelligence across both?" This shift is not semantic. It is architectural.
AI is infrastructure, not a tool
The build-versus-buy debate was designed for tools—CRM systems, analytics dashboards, marketing clouds. AI is not a tool. It is a cognitive substrate—the foundational intelligence layer on which reasoning, orchestration, and autonomous workflows are built.
Most enterprises won't create foundation models from scratch, but will fine-tune, integrate, and build on existing ones. No enterprise is going to reinvent LLMs. The foundational intelligence must be bought, just as companies once bought AWS instead of building data centers. AI is the new infrastructure, and like all infrastructure, it is a shared foundation.
Differentiation returns to the application layer
When everyone runs on the same foundation models, competitive advantage shifts to the layer above: the agentic workflows you build, how you encode domain expertise, how you orchestrate decisions, and how your systems act autonomously within context.
This layer cannot be bought. It must be built. Your institutional knowledge, sales enablement logic, your GTM playbooks, your forecasting heuristics, your internal taxonomies, your operational rhythms, your customer intelligence—these cannot be outsourced. They encode years of market learning, customer insight, operational refinement, and hard-won tradecraft. They are the fabric of the business itself.
The real enterprise decision: Which substrates do we build on?
CEOs and CIOs are no longer choosing between build versus buy. They are choosing which substrates to adopt—horizontal platforms (Scalestack, n8n, Make), domain-specific platforms (GTM Buddy for revenue activation, Brex for finance, Rippling for HR), and specialized AI (Cursor for coding, Lovable for building, Swan for Autonomous GTM)—and how to integrate intelligence across them. Horizontal platforms offer breadth and flexibility. Domain-specific platforms offer depth and velocity. Specialized AI offers autonomous execution that collapses time-to-value. The defining architectural decision of the next decade is not selecting individual platforms, but orchestrating intelligence across your entire substrate portfolio.
The substrate decisions reveal your strategic intent. Horizontal platforms make sense when your competitive advantage lives in the workflow itself—when unique processes define your business, or when you're building AI capabilities as a core competency. Vertical platforms make sense when your advantage lies in execution excellence rather than process novelty—when you need proven domain expertise embedded in the system, or when speed to autonomous operation outweighs customization.
These choices determine how fast you build, how reliably your agents behave, how much institutional context they inherit, how autonomous they can become, and ultimately, how differentiated your competitive position is. Get it right, and you compound advantage. Get it wrong, and you spend years rebuilding.
The hybrid reality
Most companies will run both—vertical substrates for functions where industry best practices apply, and horizontal substrates for workflows that differentiate you. In practice, most enterprises will operate hybrid architectures: vertical AI platforms for functions where industry best practices provide leverage (GTM, finance, customer service), and horizontal platforms for workflows that encode proprietary competitive advantage (unique pricing models, specialized underwriting logic, custom orchestration).
The strategic question isn't either/or. It's: Which functions should run on shared intelligence versus proprietary systems? This decision determines where you compound industry knowledge and where you build unreplicable moats.
AI strategy becomes a strategic flow
In the Agentic era, companies will adopt a five-phase pattern. First, buy the substrates—acquire the intelligent foundation: LLMs, orchestration platforms, agent frameworks. Second, build the autonomy layer—create agents, workflows, domain intelligence, and proprietary business logic that encodes your institutional knowledge. Third, integrate the intelligence—connect substrates, orchestrate across platforms, unify decision-making. These first three phases establish your foundation. The next two create a flywheel: fourth, operate continuously—deploy agents, measure outcomes, capture feedback loops. Fifth, adapt and expand—improve agent behaviors, extend to new domains, compound advantage. Each cycle accelerates the next.
This is not build or buy. It is buy → build → integrate, followed by continuous cycles of operate → adapt. The first three phases establish your foundation. The final two create a flywheel—each cycle accelerates the next, compounding advantage with every iteration.
The hidden costs of getting substrate choices wrong
Substrate decisions carry asymmetric risk. Choose too horizontal, too early, and your team spends 18 months building what could have been bought. The opportunity cost is steep: competitors ship autonomous workflows while you're still scaffolding infrastructure. You burn engineering resources on foundational work instead of differentiation.
Choose too vertical, too fast, and vendor lock-in limits your ability to evolve. Pre-built logic might not match your actual business model. You inherit someone else's assumptions about how your workflows should operate—assumptions that may conflict with the institutional knowledge that creates your competitive edge.
The mitigation strategy: Start with vertical platforms for proven use cases where industry best practices provide leverage. Build horizontal capabilities in parallel for workflows that encode your unique advantage. Maintain optionality through clean integration layers.
The companies that win will recognize that substrate choices are not just technology decisions—they are capital allocation decisions about where to build versus where to leverage.
The PES Framework for AI Architecture
Getting substrate choices right requires a systematic approach to AI architecture. Sustainable AI advantage requires building across three interdependent layers: People (workflows that encode institutional knowledge and expertise), Experiences (feedback loops from customer, employee, and partner interactions that continuously refine those workflows), and Systems (integration patterns that connect intelligence across your architecture and scale both). Each layer amplifies the others. Most companies optimize one layer. Winners orchestrate all three.¹
Proprietary architectures become the new moat
Understanding these architectural principles sets the foundation for a broader strategic shift. This architectural approach—building across People, Experiences, and Systems—creates a new form of competitive advantage. The companies that generate alpha won't just buy platforms and follow templates. They will design unique agentic workflows that embed institutional knowledge, build feedback loops that make their agents smarter with every interaction, create integration patterns that competitors can't easily replicate, and develop orchestration logic that reflects years of operational refinement.
This proprietary architecture layer—the connective tissue between substrates and business outcomes—cannot be bought, cannot be copied quickly, and compounds with use. This is where sustainable competitive advantage emerges in the Agentic era.
The companies that win will treat autonomy as a platform strategy, not a project
The most successful organizations won't view AI as a cost-saving automation project, a collection of isolated agents, or a patchwork of analytics tools. They will treat autonomy like the shift to cloud in 2010: a foundational capability, not a feature, infrastructure that everything builds on, a platform that grows with every use case.
Buying the right substrates and building the right autonomy layers determines how fast the business compounds. Companies that architect well will see exponential leverage. Companies that don't will face mounting technical debt and competitive disadvantage.
The Build-Versus-Buy Debate Is Over
AI collapses the traditional trade-offs between flexibility and speed, control and convenience, differentiation and standardization. You will buy the substrates. You will build the autonomy.
The question is no longer build or buy. It is: Where do you buy leverage, where do you build moats, and how do you orchestrate intelligence to compound both? This is the strategic choice of the Agentic era.
Notes
¹ The PES Framework was developed by George Alifragis through nearly two decades of operational leadership in business transformation. The framework has been successfully deployed across both public and private companies, contributing to strategic exits, and is currently applied to evaluate organizational transformation and AI architecture decisions across diverse business models.
About the Authors
George Alifragis is Senior Vice President and Head of Operating Network & Ecosystem at Metropolitan Partners Group, a New York-based private investment firm providing non-controlling growth capital to owner-operated businesses. With nearly two decades scaling public and private companies, George brings operating expertise in business transformation, strategic partnerships, and innovation-driven leadership. He has served on the Board of Desjardins and the Executive Board of the Cyber Security Global Alliance.
Karthiga Ratnam is Co-founder of Audience Haus and an Operating Expert at Metropolitan Partners Group. Pursuing her doctorate, her research and practice focus on the intersection of AI, ontology, and impact-driven category creation. Karthiga’s work helps organizations navigate the evolving landscape of technology and human understanding, turning great companies into movements people rally around.
