Large Language Model (LLM)
AI Fundamentals
A large language model (LLM) is an AI model trained on huge amounts of text to predict the next piece of text, so it can write, summarize, and answer questions.
AI Fundamentals
What Is Generative AI?
Generative AI is a type of artificial intelligence that creates new content, such as text, images, audio, video, or code, based on patterns learned from large amounts of training data. You usually guide it with a prompt, such as a question or instruction. The model generates each response rather than looking up a stored answer (though some tools search the web or your files first), so results can vary and should be reviewed before you rely on them.
Also called GenAI, Gen AI, generative artificial intelligence
Example
A five-person boutique marketing agency is preparing a seasonal campaign for a local bakery client. The account manager opens the agency’s approved generative AI assistant, adds the client’s brand guidelines and menu notes, and asks for ten email subject lines, three social captions, and an outline for a short blog post. Within minutes, the team has rough material to react to instead of a blank page.
The designer uses an image generator to explore mood-board concepts before the real photo shoot. Nothing goes out as is. A copywriter rewrites the strongest drafts in the client’s voice, checks every product detail against the menu notes, and cuts a caption that mentioned a discount the bakery never offered.
How it works
Generative AI models are trained on very large collections of examples, such as text, images, audio, or code. During training, a model learns statistical patterns: which words tend to follow others, what a product photo usually looks like, how working code is structured. Assistant products are then further trained, often with human feedback, to follow instructions and respond helpfully.
When you give it a prompt, the model produces output that fits those patterns. Most text tools built on a large language model write one small chunk of text (a token) at a time, each time predicting a likely next piece. Image generators often start from random visual noise and refine it, step by step, toward a picture that matches your description. Many tools are multimodal, meaning they can take in or produce more than one type of content, such as reading a photo and answering questions about it. Some tools also run a search first, pulling in web pages or your company’s documents for the model to draw on.
Because generation involves probability, the same prompt can return different results. The wording of your request, the examples you include, and the background information you provide all shape the output, which is why clear prompting matters.
Common business uses
Generative AI tends to be most useful where a person can quickly judge whether the output is good. Common uses include:
- Drafting: first versions of emails, proposals, job descriptions, blog posts, and ad copy.
- Summarizing: condensing long reports, meeting transcripts, or customer feedback into key points.
- Rewriting: adjusting tone, shortening, simplifying, or translating existing text.
- Brainstorming: producing headline options, campaign angles, or interview questions to react to.
- Images and visuals: concept art, mood boards, and draft graphics for a designer to refine.
- Code: spreadsheet formulas, small scripts, and plain-English explanations of existing code.
General-purpose assistants (for example ChatGPT, Claude, Gemini, and Microsoft Copilot) handle many of these tasks, and generative features also appear inside many business apps your team may already use. A practical place to start is a repetitive writing or summarizing task where mistakes are easy to spot.
Limitations to know
Generative AI produces plausible output. It does not check facts the way a careful person would, so it can state wrong information confidently, including made-up figures, quotes, or sources. This is called hallucination, and it is the main reason human review matters for anything factual or public.
- Review: treat outputs as drafts. Check facts, numbers, names, and claims before anything reaches a customer.
- Rights and ownership: rules on copyright, training data, and who owns AI-generated content vary by country and are still developing. Outputs can occasionally resemble existing work closely. Read the tool’s terms, and get legal advice for high-stakes uses.
- Data: what you paste into a tool may be stored, reviewed by the vendor, or, depending on your plan and settings, used to improve the vendor’s models. Check the tool’s data settings and your own policies before entering client, employee, or customer information.
- Bias and gaps: models can reflect biases in their training data. They may not know about events after their training data ends, or about your internal information, unless you provide it or the tool is connected to search or your company’s systems.
Generative AI vs. predictive AI
| Generative AI | Predictive AI | |
|---|---|---|
| What it produces | New content: text, images, audio, video, or code | A label, score, ranking, or forecast |
| Typical inputs | A prompt, often with examples or reference material | Structured data about a new case, such as a transaction, a lead’s details, or recent sales |
| Business examples | Drafting emails, summarizing reports, creating ad visuals | Lead scoring, fraud detection, demand forecasting |
| How you check quality | Human review for accuracy, tone, and fit | Comparing predictions against real outcomes |
| Main risk | Confident but wrong or made-up content | Wrong predictions from biased or outdated data |
The two can work together: for example, a predictive model flags at-risk customers and a generative tool drafts the follow-up message.
Further reading
- Claude 101 WorkshopA beginner-friendly, hands-on group workshop for getting practical experience with Claude, a generative AI assistant.
- AI Tool LibraryBrowse AI tools sorted by the job they do when you are ready to choose a generative tool for a specific task.
Beyond the definitions
Put AI to work for your business.
Go beyond the terminology. Explore practical AI training, workflows, and tools from SN3 Group.