AI Learning Hub › Learn AI › Self-Paced Courses › Introduction to AI › Getting Started › The AI Tool Landscape

Level 1 of 4 · Getting Started · Lesson 1 of 6

The AI Tool Landscape

The main kinds of AI tools, and how providers, models, applications, features, and accounts differ.

About 15 minutes

Skip to lesson

By the end of this lesson you should be able to

  • Place a tool in one or more families: assistants, research tools, AI inside everyday software, legal systems, media tools, and tools that act
  • Tell apart a provider, a model, an application, a feature, and an account
  • Read a product description and name one thing it does not establish about your own access or use

Start with the task, not the logo

Three ordinary tasks at the law school: a student wants help understanding a dense public reading before class; a staff member needs a short announcement for a public lunchtime talk; a 1L wants to practice explaining a case out loud before being called on. All three could involve AI, and none of them starts with a brand. The useful questions are what capability the task needs (explaining, drafting, conversation, feedback on speaking) and which tool you are permitted to use for it.

This lesson gives you a map of the kinds of AI tools you will meet and the words people use to describe them. AI here means computer systems that do tasks we associate with human thinking, such as understanding and producing language. Most of the tools in this level are built on generative AI: systems that produce new text, images, audio, or code. Later levels explain how they work; for now, what matters is what they are for.

Six overlapping families of tools

These families overlap. One product can belong to several, and products move between them as features are added.

General-purpose assistants
Chat-based tools that explain, draft, summarize, and answer follow-up questions across almost any subject. Examples at Stanford include ChatGPT Edu, Claude for Education, Gemini, and the Stanford AI Playground (Lesson 2).
Research and source tools
Features and products built around finding or working with sources: web search inside an assistant, longer research modes, and notebooks that work from documents you choose (Lesson 3).
AI inside familiar software
AI added to tools you already use, such as a word processor, email, or a meeting app. Whether it is switched on depends on your organization’s license; at Stanford, for example, some of these built-in features are licensed separately or turned off.
Specialized legal systems
Tools designed around legal research, legal documents, or advocacy practice, such as Lexis Protégé, Westlaw’s AI features, Harvey, and CICERO (Lesson 5).
Creative and media tools
Tools that make or edit images, audio, or video, or describe and transcribe them (Lesson 4).
Tools that work with files, code, or other apps, and can act
Features that run code, open files, connect to other applications, or carry out steps such as sending or scheduling. These raise questions about permissions, not just quality (Lessons 2 and 6).

Five words that are easy to mix up

Provider
The company or institution that offers something. OpenAI, Anthropic, and Google are providers; so is Stanford, when it offers a service to the community.
Model
The trained AI system that produces the output. A provider may offer several models, and one model can sit behind many applications.
Application or service
The product you actually open: a website, an app, or a feature inside other software. It decides what the model can see and do: which files, which tools, which settings.
Feature
A particular capability within an application, such as web search, file upload, or voice.
Account
The identity you sign in with, such as a personal account or a Stanford one. An institution-managed account usually belongs to a workspace: the organization’s environment, where administrators decide which features, apps, and settings are available. The same product can behave differently under a personal account and an institutional one, and features can differ by plan. A project is something else again: in many applications, a container that groups related chats, files, and instructions for one piece of work. Product labels vary, so check what a given product means by each.

Brand names blur these lines. “Claude” and “Gemini” each name both a family of models and an application, and “ChatGPT” names an application that offers several models. When someone says “I used Gemini,” it is worth asking which application, which account, and which features.

Two more terms come up often. A connector lets an application reach another service, such as a cloud drive, so it can read (or sometimes change) material there. An agent is an AI system that works through several steps toward a goal, choosing actions along the way, sometimes including actions outside the chat. Connectors are introduced in Lesson 2; both are treated fully in the Advanced level.

Worked example: one model, two applications

A constructed example. The model, products, and settings are fictional.

A fictional model called Orion is available in two places:

  • Northstar Chat, the model maker’s own app: web search is available, you can upload files, and it can remember preferences between conversations. Anyone can sign up with a personal account.
  • CampusDesk, a university service: the same Orion model, but no web search; it can search only a collection of course readings the university has loaded, and you sign in with a university account.

Asked about a news event from last week, Northstar Chat could try a web search. CampusDesk has no web search in this scenario, so it has no live route to the news, though you could still paste in the text of an article yourself. Asked about an assigned reading, CampusDesk can retrieve it from the course collection the university loaded. Northstar Chat needs the reading supplied, by upload or pasted text, or found through a successful search if the reading is publicly available; nothing in the scenario says it already has the text, and having a search feature is not the same as finding the right source. Same model, different applications, different routes to information. The model name alone does not tell you what a tool can reach, which is why “which model?” is rarely the only question worth asking.

Try it: read a product card

A fictional product card. Nothing here describes a real product.

Lumen Desk, from Brightline Software. Runs on the Orion model from Northstar AI. Features: web search (open web only; no subscription databases); upload PDF and Word files; voice input. Available through organizations that have signed up with Brightline; file upload is switched off on personal accounts.

  1. Name the application provider, the model provider, and the model.
  2. List the features, and say which one comes with a limit on its sources.
  3. What account condition does the card describe?
  4. Name one thing the card does not tell you that you would need to know before using Lumen Desk for real work.
Answers and discussion
  1. Application provider: Brightline Software. Model provider: Northstar AI. Model: Orion. The company whose name is on the app is not the company that made the model.
  2. Features: web search, file upload (PDF and Word), and voice input. Web search is limited to the open web, so it cannot reach subscription databases.
  3. It is offered through organizations, and file upload is off for personal accounts. Someone using a personal account would not get the same product.
  4. Several answers work. The card does not say whether your organization offers it or has switched features on; whether your information and your intended use are permitted in it; whether its answers show sources you can check; or how well it performs on your kind of task. A feature list describes what the product is designed to do, not what your account can do or how reliable the results are.

Before you try anything

  • Practice with fictional material, or public material you are permitted to use. A fictional scenario, your own notes, or a public reading whose terms allow this use. Being publicly available does not by itself mean material may be uploaded or reused for any purpose.
  • Check the rules for the task. For coursework, your instructor’s AI policy comes first; Beginner Lesson 5 explains the Stanford Law School default and how to think about information.
  • Use the Stanford access route before giving a tool any work information. The Hub’s AI Resources page lists the tools available to the SLS community and where to request access.

Throughout this level, remember three things. These are tools. A tool that writes fluently is not thereby a person, an expert, or an authority. And you stay responsible for how you use it and for checking what it gives you.

Ask four questions of any tool. Which application is this? Which account am I using? Which features does it actually have for me? And is this use, with this information, permitted? A product name alone answers none of them.

Where to go next