TensAI maps the territory of enterprise AI. This is not a vendor pitch, and it is not a framework tutorial. It is a structured, opinionated body of work on building AI properly across an organisation — the kind of work that requires decisions about foundations before features, and coherence before capability.

The taxonomy organises 18 canonical concepts across three pillars: Groundwork, Operating Model, and Craft. Groundwork establishes what has to be true before AI can be trusted — data quality, governance, security posture. Operating Model addresses how AI is introduced, scaled, and sustained across teams and processes. Craft concerns the deliberate discipline of building AI systems that behave as intended. One governing principle runs through all three: whole before parts. Architecture precedes implementation. Context precedes tooling.

TensAI is vendor-neutral and standards-first. The protocols matter more than the products. MCP, Agent Skills, A2A, OpenTelemetry — these are the connective tissue of a serious AI practice. Any organisation that anchors its AI programme to a specific vendor's abstraction layer will eventually pay the cost of that decision. TensAI takes the position that durable AI capability is built on open standards and rigorous thinking, not on API convenience.

The work is built for three kinds of readers: CTOs who are making structural decisions about where AI fits in their organisation, platform leads who are responsible for scaling AI safely across teams, and practitioners who are building the systems and need a shared vocabulary for the hard problems. If you are any of those, this is for you.