If we couldn't handle the rigid logic of ERPs, we are entirely unprepared for the probabilistic chaos of Artificial Intelligence.
There is an old, somewhat cynical investment strategy in the technology sector: short any company announcing a large-scale Enterprise Resource Planning (ERP) implementation.
For the last two decades, we have watched organizations struggle to define their operations clearly enough to stand up these massive systems. ERPs demand a ruthless precision—a distinct "physics" of how a business operates—that most companies simply cannot articulate.
If we couldn't handle the rigid logic of ERPs, we are entirely unprepared for the probabilistic chaos of Artificial Intelligence.
AI transformation is not just a software upgrade; it is an architectural overhaul of the enterprise, demanding a total re-definition of roles, data pipelines, and organizational charts. We are standing at a precipice where the market is about to split into three distinct tiers. And for the Product Managers watching this unfold, the message is stark: evolve technically, or face obsolescence.
This is the rarest tier. These are the companies pursuing full-blown implementations of their own Large Language Models (LLMs). They aren't just using AI; they are re-architecting their entire business model to be "AI-first." They have solved the "definition problem" that plagued ERPs.
This group will survive by partnering. They will leverage providers like Google or OpenAI, investing heavily in Retrieval-Augmented Generation (RAG) to pipe proprietary enterprise data into LLM contexts. This won't necessarily transform their business model, but it will successfully augment existing roles and improve efficiency.
This is where the majority of companies will land, and where they will fail. These organizations will simply "throw everything into the LLM," hoping for magic.
The result will not be differentiation. It will be a regression to the mean.
When you rely on generic models without strict architectural constraints, the output is not insightful—it is statistically average. We are identifying two primary symptoms of this "AI Drift":
If the business risk is homogenization, the career risk for Product Managers is extinction. The rise of "vibe coding"—where natural language prompts allow engineers to build complex systems rapidly—is lowering the barrier to entry for software creation.
// The Technical Imperative
You must understand the difference between deterministic systems (databases) and probabilistic systems (LLMs). If you cannot understand why an LLM outputs a "false range" or struggles with Markdown, you cannot manage the product risk.
The warning signs are clear. If your organization cannot define its business operations to the level of detail required for an ERP, it cannot support an AI transformation.
The winners of the next decade won't be the ones who "underscore the importance of AI". They will be the ones who understand the machinery well enough to take it apart and rebuild it.