AI Glossary of Terms

nAItion & ImaginAItion Network Edition

Plain language. No hype. If a term is specific to the nAItion network, it's marked [nAItion].

A

Agent A software entity that takes action on your behalf. Not a chatbot that answers questions — an agent that does things: books meetings, drafts proposals, reviews invoices, flags risks, routes tasks. Your ImaginAItion OS staff is made up of agents. They work while you sleep.

Agentic Capital [nAItion] The ImaginAItion Capital approach to small business lending. Because the OS has real-time access to your books, receivables, and run rate, underwriting is automated. Pre-approved members see a "Request Capital" button in their workspace. The agent underwrites against ground truth — no statements required, no self-reported numbers. Capital deploys via ImaginAItion Pay.

AI (Artificial Intelligence) Software that performs tasks typically requiring human-level judgment — reading, writing, pattern recognition, decision-making, conversation. The word is overused. In the nAItion context, AI is infrastructure, not a feature. Every agent, every module, every insight runs on it. We don't call things "AI-powered" — everything here is.

AI-Native Built from the ground up with AI as the operating layer, not added on afterward. The difference between a building designed for electricity and one that had wires stapled to the walls later. ImaginAItion OS is AI-native. Most of what your competitors sell you is the latter.

Autonomy (AI) How much authority an AI system has to act without human confirmation. Higher autonomy = more actions taken independently. Every agent in the nAItion network has an explicit autonomy level — and every action leaves an audit trail. Trust is earned, not assumed.

B

Benchmark A standardized test used to compare AI model performance. Models are graded on benchmarks for reasoning, coding, instruction-following, math. Useful for routing decisions — the right model for the right task. Not a measure of whether a model is right for your business.

Brand Voice The specific rules and constraints that govern how a brand speaks. Tone, vocabulary, banned phrases, outcome-first framing. In the nAItion, every member's brand voice is governed in ImaginAItion OS's Brand Hub — and the network's own voice (what goes on the nAItion site, what agents say to your customers) has its own doctrine.

C

Closed Economic Loop [nAItion] The structural design of the nAItion. Sub-co revenue funds the membership pool. Members earn on every dollar that stays in the network. The loop closes when members stop sending money to outside vendors and route it through the network — keeping the economics inside.

Context Window The amount of text an AI model can hold in "working memory" at once. Longer context windows mean the model can read more of your business history, your contracts, your communications, before it responds. Most consumer AI has a short memory. The nAItion's agents are built to use deep context on your specific operation.

Covenant [nAItion] The all-in membership tier, coming soon. High usage, high commitment, mutual obligation. The platform commits back to Covenant members with service levels and accountability that lower tiers don't carry.

D

Deep Learning A type of machine learning using layered neural networks to recognize patterns in data — images, language, audio. The technical foundation behind most modern AI capabilities. Not a metaphor. An actual architecture.

Distillation Taking a large, expensive model and training a smaller, cheaper one to replicate its outputs on a specific task. The smaller model is faster and cheaper but good at the narrow job. The nAItion routes work to distilled models where appropriate — the expensive model for hard calls, the distilled one for volume.

F

Fine-tuning Training an existing AI model on new, specific data to make it better at a particular task. A general model fine-tuned on your industry's vocabulary and your company's history becomes dramatically more useful than a general model answering cold. Fine-tuning is how the nAItion's agents develop operational depth over time.

Founding Believers [nAItion] The nAItion's earliest backers. Investment opportunities are available: see the investor page at investors.imaginaitioninc.com.

Foundation Models Large AI models trained on broad datasets — the base layer everything else is built on. GPT, Claude, Gemini, Llama are foundation models. The nAItion's agents run on top of them. Foundation models are infrastructure, not product — like the electrical grid. What matters is what gets built on it.

G

Generative AI AI that produces new content — text, images, audio, video, code — rather than just classifying or predicting. The technology behind large language models, image generators, voice synthesis. What most people mean when they say "AI" in 2024–2025.

Guardrails Constraints built into an AI system to prevent specific behaviors — fabricating information, making decisions outside its authority, acting outside its defined role. Every agent in the nAItion network has constitutional guardrails. Guardrails limit the damage. They are not trust — trust is built with an agent over time, not installed around it.

H

Hallucination When an AI model generates confident-sounding information that is incorrect or fabricated. A real failure mode. The nAItion's agents are built to flag uncertainty rather than invent — and audit trails exist so fabrications can be caught and corrected. No system is immune. Honest architecture catches it.

Human-in-the-Loop A design where a human reviews and approves AI actions before they execute. Higher-stakes or irreversible actions require it. Lower-stakes, routine actions run without it. The nAItion's agent autonomy system is explicit about which is which.

I

Inference The act of running an AI model — asking it a question, giving it a task, having it generate output. Training is expensive and rare. Inference is what happens every time you use the product. Inference costs are what you're paying for in AI compute.

Integration Connecting two systems so they share data and actions. CRM integrated with email means your email history shows up in your customer record. Most small business software integrates poorly. ImaginAItion OS is built integration-first — the data flows so agents can act on ground truth, not stale exports.

L

Large Language Model (LLM) An AI model trained on massive text datasets to understand and generate language. The technical category that includes GPT, Claude, Gemini, Llama. LLMs are the reasoning engine inside most AI agents today. Not magic — pattern recognition at enormous scale.

Latency How long it takes an AI model to respond. Fast models have lower latency. Slower, more powerful models have higher latency. The nAItion routes tasks by urgency — real-time work gets fast models, complex reasoning gets the slower, deeper ones.

M

Membership Pool [nAItion] The economic heart of the nAItion. Sub-co revenue flows in. Members earn out. The pool is the structural mechanism that makes co-ownership real — not a metaphor, not a loyalty program. A literal revenue-sharing instrument tied to network performance.

Model An AI system trained to perform specific tasks. Different models have different strengths — some are better at reasoning, some at speed, some at code, some at conversation. The nAItion routes work across multiple models depending on what the task needs. No single model is best at everything.

Module A discrete, reusable unit of functionality within ImaginAItion OS. A shipping module, a contract module, a scorecard module. Modules are built once and deployed across members in the same vertical. The more members in a vertical, the better the module gets.

Multi-agent A system where multiple AI agents work together, each handling a specific role, coordinating on a shared task. The ImaginAItion OS staff model is multi-agent: your CFO agent, CMO agent, CEO agent, Delivery agent — each specialized, all coordinated.

O

Orchestrator An AI agent whose job is coordinating other agents — routing tasks, synthesizing results, making sure the right work goes to the right place. In the nAItion network, the orchestrator is the agent above all other agents: it routes the work, holds the context and hands each task to the right specialist.

P

ImaginAItion OS [nAItion] The operating system the nAItion runs on. Purpose-built for small businesses. Every suite (creAIte, mAIrket, operAIte, delAIver, servAIce, portfolAIo), every agent, every module connected under one login. Not a CRM. Not a project tool. The whole operating layer.

Prompt The input given to an AI model — a question, an instruction, a task description. Quality of prompt affects quality of output. The nAItion's agents are pre-prompted with your business context so you don't have to explain yourself every time.

Prompt Engineering The practice of designing inputs to AI models to reliably produce high-quality outputs. Less art, more craft. In production AI systems, good prompt architecture is engineering work, not improvisation.

provenance mark [nAItion] The capital AI inside ImaginAItion's own names — ImaginAItion, ImaginAItion OS and the nAItion — and in the apps sold on their own, like pAIrtner, portfolAIo and fAImily. It's not a stylistic choice. It's a hallmark. Every product bearing this mark was built inside the ImaginAItion network.

R

RAG (Retrieval-Augmented Generation) A technique where an AI model retrieves relevant information from a knowledge base before generating a response. Instead of relying only on what it learned during training, it looks things up first. How agents give accurate, current, business-specific answers rather than generic ones.

Reasoning Model An AI model with extended capacity to think through multi-step problems before answering — rather than producing an immediate response. Slower. More expensive. More accurate on hard problems. Used for complex underwriting, strategic analysis, legal reasoning.

S

Sub-co [nAItion] A sub-company within the nAItion network. Each is a separate LLC, operated by a domain expert, serving the network's members at network-negotiated rates. Sub-co revenue funds the membership pool. Examples: Capital (lending), Assurance (insurance), Legal (compliance), Supply (sourcing), Creators (brand deals).

Substrate The foundational layer a system runs on. In AI terms, the data, memory, and context that an agent draws from before responding. The nAItion's agent substrate includes your business history, your contracts, your financial position, your brand voice — so agents operate from ground truth, not approximation.

T

Token The unit AI models use to process language. Roughly 3/4 of a word. Models have context windows measured in tokens. Pricing is per token. Performance is affected by token load. Not something most users need to think about — relevant when understanding AI infrastructure costs.

Training The process of teaching an AI model by exposing it to large datasets and adjusting its internal parameters to recognize patterns. Expensive. Rare. Most AI usage is inference (running a trained model), not training. Fine-tuning is a lighter version of training on specific data.

Twin (AI) A digital representation of a person built from their voice, writing, decision patterns, and history — enabling interaction with a version of that person's knowledge and perspective. The nAItion's Our Roots product uses this architecture for family preservation. Members can capture enough of a person that family can still interact with them after they're gone.

V

Vector Database A database that stores information as mathematical embeddings (vectors) rather than rows and columns — enabling semantic search. "Find me things similar to this" rather than "find me things that match exactly this keyword." Used to power memory in AI agents so they can retrieve relevant context rather than requiring exact phrase matches.

Vertical A specific industry or market category. Fitness, hospitality, recovery, legal services. The nAItion builds vertical-specific agent stacks — the same operating architecture, tuned to the language, metrics, and workflows of each industry.

W

Workflow A sequence of steps — automated, human, or mixed — that produces a business outcome. AI-native workflows replace linear human-approval chains with agents acting in parallel. The goal is not removing humans; it's removing the friction between good decisions and their execution.

Built using imaginAItion. A nAItion company. Believe · Build · Evolve