Glossary
Plain definitions of the terms that come up when you build, design and launch AI products and ventures. Each one opens with a two-sentence definition, then says how we use it, with an example from our own work or method.
A
- Agentic workflow
- An agentic workflow is a process in which an AI model plans and carries out several steps on its own, using tools such as search, a database or other software, rather than answering a single prompt.
- AI-native
- An AI-native product is built around what a model can do, rather than having AI added to a product designed without it.
B
- Brand OS
- A brand OS is the operating system for a brand: its strategy, verbal identity, visual identity, motion and governance, documented so that anyone, including AI tools, can apply it consistently.
D
- Design partner
- A design partner is an early customer who works closely with a team while a product is being shaped, trying early versions and giving feedback in return for influence and often early access.
- Design system
- A design system is the shared set of components, rules and decisions a product's interface is built from: colours, type, spacing, patterns and the code that implements them.
E
- Evaluation set
- An evaluation set is a collection of real inputs, each paired with what a good output looks like, used to score an AI system before it's released.
F
- Fine-tuning
- Fine-tuning is training an existing AI model further on your own examples, so it adapts its behaviour to a particular task, style or domain.
G
- Guardrails
- Guardrails are the rules and checks around an AI model that keep its behaviour within safe, useful limits: what it may answer, what it must refuse, what data it may touch and when it hands over to a person.
H
- Hallucination
- A hallucination is an answer from an AI model that sounds confident and is wrong: an invented fact, a source that doesn't exist, a number with nothing behind it.
- Human in the loop
- Human in the loop means a person reviews, approves or corrects an AI system's work at defined points, rather than letting it act entirely on its own.
I
- Ideal customer profile (ICP)
- An ideal customer profile describes the kind of customer a product is best for: who they are, what they're stuck on, and why they'd buy.
J
- Jobs to be done
- Jobs to be done is a way of understanding demand by asking what progress a customer is trying to make, and what they "hire" today to make it.
K
- Kill criteria
- Kill criteria are the results that would make you stop a product, a venture or a test, written down before it starts.
M
- Machine legibility
- Machine legibility is how easily AI systems, search engines and agents can read, understand and accurately describe a company or product.
- Minimum viable product (MVP)
- A minimum viable product is the smallest version of a product that can test whether people want it.
P
- Product-market fit (PMF)
- Product-market fit is the point at which a product meets a real need well enough that the right customers keep using it, come to rely on it and pay for it without being pushed.
- Prompt injection
- Prompt injection is an attack in which text given to an AI model, such as a message, a web page or a document, contains instructions that try to override what the model was told to do.
R
- Retrieval-augmented generation (RAG)
- Retrieval-augmented generation, or RAG, is a way of having an AI model answer from your own documents: the system first finds the passages most relevant to a question, then gives them to the model to write the answer from.
S
- Soft launch
- A soft launch opens a product to new users without an announcement, to check that the setup works for people you didn't hand-pick before telling everyone.
- Structured data
- Structured data is information added to a web page in a standard format, usually JSON-LD using the Schema.org vocabulary, that tells search engines and AI systems exactly what the page describes: an article, an organisation, a question and its answer.
- Stuck-to-Scale
- Stuck-to-Scale is Graylemon's model of the six stages a venture moves through: Ideate, Define, Build, PMF, Market and Grow, each with the one question it has to answer before the next can start.
V
- Venture studio
- A venture studio builds companies repeatedly, supplying the idea, the team, the product and the early resources itself, rather than investing in founders who arrive with their own.
- Vertical AI
- Vertical AI is an AI product built for one industry or kind of business, shaped around its workflows, data and rules, rather than a general tool for everyone.
W
- White-label
- White-label work is built by one company and delivered under another company's brand, so the end client sees only the brand they hired.
- Willingness to pay
- Willingness to pay is the evidence that customers will pay for a product at a given price.