The Architecture of Advantage: Five Enterprise Tech Principles for the AI Era

Kelly Coutinho

September 17, 2026 • 5 min read

---

Since January 2026, 67% of our conversations with supply chain and operations leaders have been about the technology stack. These discussions, whether focused on spend, vendor assessment, or new AI capabilities, all point to a broader, more urgent theme:  the need to match operational reality with tech choices to ensure competitive advantage in the AI era.  

The ground has shifted. Now, tech spend has a much closer relationship with operational strategy. As an operations leader at a multinational food and beverage company recently told us, “If you’re going to use more agentic AI, you’ll transmit more data, you’ll run out of your bandwidth, then you’ll have to upgrade your capacity.”  The link between AI, spend, and infrastructure capacity is now direct and unavoidable.  

This convergence is also reshaping workflows. For instance, a machine participating in a decision needs deep context  – relationships, permissions, and business rules – that transcend traditional application and functional silos. To this point, in recent analysis of one end-to-end workflow at a consumer goods company, we found that 55% of the underlying capabilities did not belong exclusively to a single function. Intelligence must now flow across the very boundaries enterprise tech was built upon.  

In our Signal a few weeks ago, we shared why addressing this new reality by matching operational reality with technology choices requires answering two questions:   

• What specific work are you trying to change?   
  Competitive advantage starts with the work. Where can technology materially improve how the business operates, and which capabilities become more valuable as they are reused elsewhere? This tells us what is worth creating.  
•  What architecture do you already have?   
  Architecture determines whether capabilities can travel, adapt, and scale while preserving the economics, control, and optionality the enterprise needs. This determines whether advantage can compound.   

In this Field Note (and over the next few weeks), we’re spotlighting the five Zero100 principles that provide a durable framework for answering these questions.  

The Five Principles of Modern Enterprise Technology  

1. Enterprise technology is only valuable when it changes how work gets done  
   There is no shortage of viable technology options. The more useful starting point is the work: where judgment is being applied, which decisions determine the outcome, and what would materially improve as a result. This can sometimes require new technology, but it may instead require making the existing estate work better together. As AI makes it easier to add intelligence into more parts of the business, the bar needs to shift from “the technology works” to “the work changes because of this technology.”   
2. Competitive advantage comes from capabilities that compound through reuse  
   Many of the capabilities underneath functional workflows are shared. For example, demand history can inform inventory, manufacturing, and capacity. This means that if every function repeatedly engineers its own data, context, intelligence, and operating logic, value stays local. Designing selected capabilities for reuse allows the same investment to keep creating value elsewhere.  
3. Architecture determines what scales  
   AI has made experimentation considerably easier, but enterprise scale is still difficult. One pharmaceutical organization we worked with started with a PLM upgrade. Once we decomposed the requirement, product changes needed to translate into planning decisions, with intelligence spanning two platforms and two parts of the organization. This is an example of how the technology question changes to: where should the capability live, what warrants internal ownership, and what can be reused elsewhere?
4. Intelligence runs through the stack  
   Models and agents dominate much of today’s AI conversation, but enterprise intelligence depends on what surrounds them. To be useful in an enterprise, AI needs the right data, business context, rules, and constraints around it.  As decisions move, that context builds. Models will improve, and providers will change. The data, relationships, operating context, and judgment that explain how an enterprise works are likely to be far more durable.  
5. Every layer requires an explicit ownership choice  
   Build vs buy is becoming too coarse and reductive a binary for the systems enterprises are assembling. A workflow might combine an external model and cloud infrastructure with enterprise-owned data and proprietary operating logic before landing in a third-party application. There is no reason each layer should carry the same ownership model.  

Comparing our data and analysis across three consecutive quarters, 67% of companies are choosing to partner with tech providers (Q2 2026) vs 40% in Q4 of 2025. It’s a directional signal: greater partnership does not necessarily mean relinquishing control. Our view? Own what differentiates and compounds, partner where external capability accelerates, and allow commodity layers to commoditize.   

What Comes Next  

These principles will not produce the same technology strategy for every organization. Nor should they.  

We’ve recently seen two organizations pursue similar ambitions around connecting signals into planning decisions and arrive at very different answers. One, working with constrained capital and a fragmented legacy estate, connected selected capabilities while preserving much of its existing technology. Another, already undertaking significant transformation, chose to internalize more of its shared data, semantics, and intelligence capabilities.   

Technologies will change. But our focus is on what should endure: the capabilities that create advantage and the architecture that allows them to compound.