Industrial AI terminology guide — extended companion

Industrial teams are flooded with overlapping labels—LLM, agent, agentic, RAG, multi-agent—often used interchangeably in decks and RFPs. When those words mean different things to engineering, procurement, and legal, you mis-size risk, cost, and accountability before the first production gate.

This document is the extended companion to AI²’s public terminology guide. It sets out a shared, vendor-neutral vocabulary co-developed by Vlad Larichev and Alexey Samoshilov for the AI² network: what each term implies in industrial settings (plants, regulated products, capital projects), not only in consumer-style chat or coding assistants.

The spine of the work is a five-layer model (A through E)—from foundation models and LLMs (A), through retrieval and grounding (B), agents and tool use (C), agentic orchestration (D), to economics, governance, and human-in-the-loop design (E). The PDF walks each layer so readers can map vendor claims, internal roadmaps, and diligence questions to the same ladder.

Compared with the free web article, the PDF goes deeper: longer definitions, more worked language on procurement and readiness, a practical “vendor conversation decoder”, and literature-backed references with an expanded bibliography for teams that need citable sources in reviews and steering packs.

This edition refreshes the narrative and reference set while keeping the same A–E framework as the site. Use it in architecture reviews, vendor sessions, and training so everyone—from shop-floor digital leads to the C-suite—can argue about substance, not slogans.

Download the PDF below. The companion article From LLM to Agentic AI: a practical guide to the terms that matter covers the same themes in a shorter, web-native format and links to this file.