Why Semantic Operations?

What is SemOps? introduces the framework and its core thesis: the conditions that make AI work well are the same conditions that make organizations work well. This page elaborates on the problem we're solving and how I arrived at this approach.

AI Is Analytics

If you dive into the details, AI is fundamentally a form of analytics. LLMs, agents, RAG pipelines — they process data, identify patterns, and generate insights. The technology is impressive, but the underlying operation is what analytics has always done: convert structured information into decisions.

Most people have experienced this with a good chatbot. It understands context, connects ideas, and produces genuinely useful answers — it can feel like magic. That works because the conversation is coherent: one person, one topic, shared context. Where I think AI shines even more is agentic coding — code has inherent structure, rules, and clear boundaries, and AI thrives in that environment. AI performs well wherever meaning is explicit.

The supporting technologies tell the same story. Retrieval Augmented Generation (RAG) is mostly non-AI primitives — search indexing, querying, and data pipelines with a model on top. Agentic platforms are largely governance — controlling what agents can do, with what data, under what rules. The AI model itself is a powerful but relatively small part of a much larger system, and that system determines how well the model performs.

The Coherence Gap

Achieving similar coherence across analytics and enterprise systems — where most organizational work happens and decisions are made — is a genuine challenge. It isn't a new one that AI created; it's the core knowledge and coordination problem these systems have always presented.

Data systems are complex and cross-cutting, and often misunderstood by stakeholders for exactly those reasons. Teams own different components, governance is fragmented, and the reports and dashboards that consume them can be easily misinterpreted or quietly wrong. Architecture accumulates rather than being designed — systems get built, integrations get added, and the result rarely reflects the actual business domain. Technical systems and organizational structures drift apart over time.

These are organizational challenges, not purely technical ones. The skills, ownership, and shared understanding required to maintain coherent systems don't emerge on their own. They require deliberate investment in how people across roles — technical and non-technical — think about and manage their systems together.

AI Raises the Stakes

AI doesn't fix the coherence gap. It makes it more visible and more consequential.

AI is expanding what counts as "data" — documents, conversations, decisions, code — on top of data systems that are already difficult to manage, are underresourced, or don't exist. Every gap in traditional data management becomes a failure mode for every agent interaction. When an AI system inherits conflicting definitions, unclear ownership, or inconsistent structures, it does no better than a human working with the same limitations.

At the same time, AI expectations often outpace reality. ROI for organizational AI integration is not meeting expectations, and the instinct that individual benefits will scale to cross-cutting teams and systems runs into the reality that AI needs coherent inputs to produce coherent outputs.

But AI also provides powerful tools to address the gap. With the right conditions in place, AI can accelerate the work of building and maintaining coherent systems — profiling data, mapping relationships, encoding business rules, and identifying inconsistencies that would take humans weeks to find. The technology that exposes the problem is also part of the solution.

The Conditions Thesis

This leads to the core observation behind Semantic Operations:

The conditions that make AI work well overlap significantly with the conditions that make organizations work well, with or without AI. SemOps is built around this overlap. Clear definitions, consistent structure, explicit governance, domain-aligned architecture, shared understanding across roles — these are operational fundamentals that improve decisions and reduce confusion whether or not AI is in the picture.

Achieving the right conditions can be difficult. But part of the playbook includes using AI to do this as well, and the benefits start from day one. Every improvement in coherence — a clearer definition, a better-governed data system, architecture that reflects the business domain — pays off immediately in human decision-making while simultaneously making AI more effective.

There's a compounding effect: better conditions enable better AI, which helps create better conditions. That flywheel is what Semantic Operations is designed to set in motion.

So Is It Worth It?

The problem space above may seem daunting, but these problems exist whether they're addressed or not, and there is real upside. I maintain that AI is analytics, but it is massively powerful analytics. I think it can genuinely change how work gets done, accelerate decisions, and multiply the efforts of those willing to adapt. The Semantic Operations Framework is one approach to creating the conditions where that potential becomes real.

The real question isn't "how do we integrate AI?" It's bigger: now that AI is here, what's possible if we're willing to really think it through?