Glossary

An A-to-Z of the AI-transformation vocabulary an owner runs into across this wiki.

Owner at a reading desk with an open reference book, crisp navy-tabbed sections, gold accent on the spine.

Recurring terms across the wiki. Where a term has its own page, follow the link for the full treatment. Definitions here are deliberately short.

A

Agent / agentic AI: AI that does not just answer but carries out multi-step work, taking actions toward a goal rather than producing a single response.
Agent Opportunity Analysis: Ranking where in your business an agent could actually be applied for real return. See Agent Opportunity Analysis.
Agentic Business Blueprint: The discovery process that maps how your company really runs and produces a ranked, ROI-backed plan for transformation. See The Agentic Business Blueprint.
AI psychosis: A term coined by Box CEO Aaron Levie for leaders who, being too far from the work, mistake a clean demo for a solved problem. See AI Psychosis in the C-Suite.
AI valuation premium: The materially higher valuation buyers assign to companies that genuinely run on AI versus comparable peers that do not. See The AI Valuation Premium.

Benchmark: A standardized test of model capability. A high benchmark score is not the same as capability in your business, since benchmarks often test narrow or already-seen problems.

Happy path: The version of a demo where nothing ever goes wrong, which has never survived contact with a real user.
Hallucination: When a model confidently produces something false or invented; a core reason AI cannot be left unsupervised on high-stakes work.

Multiple: The factor applied to a company's earnings to arrive at its value. Genuinely embedded AI is now one of the variables that move it. See Why AI Now Sets Your Ceiling.

Operational maturity: How well a company runs as a system rather than through one person's head; the real thing buyers and operators value.
Organizational Truth Repo: The checked-in source of truth (brand, vision, priorities, SOPs) that makes your own AI generate on-brand, company-aware work. See The Organizational Truth Repo.
Owner dependence: The structural cap that exists when the company, or its AI leverage, lives only in the owner's head. See Owner Dependence.

ROI (return on investment): The honest test for any AI initiative: does it solve a specific, high-impact problem with measurable return, or is it hype-driven.
Runtime: The environment a model actually executes in; part of the substrate you should control. See Own the Substrate.

SOPs (standard operating procedures): Documented processes that let a business run without the owner, and a real operational asset. See SOPs as an Asset.
Substrate: The full reasoning surface under your AI: weights, prompts, version, tokenizer, runtime, and hardware. See Own the Substrate.
Sycophancy: The well-documented tendency of language models to agree with the user rather than challenge them, which degrades a leader's judgment. See AI Psychosis in the C-Suite.

Templated solution: A generic, off-the-shelf AI offering that encodes the average company. The tell that a provider has not studied your business. See There Is No Templated Transformation.
Transferability: The degree to which an operation, including its AI leverage, can be handed to someone else and keep running. See Documentation Equals Transferability.
Transformation flood: The wave of AI-transformation partners arriving, with incentives that do not perfectly match yours. See The Transformation Flood.

Weights: The trained parameters that are the model itself; owning them is part of owning the substrate. See Own the Substrate.