HASZB_AIHASZB_AI

Search across courses, lessons, glossary terms, prompts and tools.

Stage 5 · AI Tools

The tool landscape

A map of the categories, so you can tell which kind of tool a task needs before you start comparing products.

6 min read

In this lesson

  • Name the main categories of AI tool and what each is for
  • Match a task to a category before comparing products
  • Understand why most tools are built on a handful of models

The tool market changes faster than any list can track. Categories change slowly. Learn the categories and you can place a new product in about a minute, which is more useful than memorising names.

The categories

General assistants. A chat interface over a general model. Broad, conversational, good for reasoning, drafting and explanation. The default starting point for most tasks, and often sufficient.

Coding tools. Editor integration, awareness of your repository, and the ability to run and iterate. The important difference from a general assistant is context: it can see your codebase, so its suggestions fit your actual project.

Research and document tools. Built around retrieving from a specific corpus — your documents, or the web — rather than model memory. This is the category to reach for when accuracy against sources matters more than fluency.

Image, audio and video. Generation and editing in other media. Different underlying architectures, different failure modes, and the most active legal and licensing questions.

Automation platforms. Connect services and put AI decision steps inside a workflow that runs without you. Covered properly in a later stage.

Agent platforms. Tools for building systems that plan and take actions, rather than answering a question. The subject of stages six and nine.

Why so many tools feel similar

A large share of products are built on a small number of foundation models accessed through an API. That is not a criticism — it is the normal shape of the market — but it explains something useful.

If the engine is shared, the product is everything around it:

  • What it can see. Your files, your repo, your email, a specific corpus.
  • Where it sits. In your editor, your browser, your document, your chat.
  • What it does with failure. Does it say it does not know, or invent something?
  • What it does with your data. Retention, training, and where it is processed.

Those four questions separate tools far better than any claim about capability.

The honest default

For most people, most of the time, a single general assistant covers the majority of tasks. Specialised tools earn their place when you have a specific, repeated task where the general tool's lack of context is the bottleneck.

Adopting six tools before you have a repeated task is a way of spending money and attention to feel current. Start general, add specific tools when a real friction appears.