omol Join

Tutorial

AI Prompting Masterclass: Full Course with Andrew Ng

You can vote once per content item.

AI has changed dramatically since the first generation of consumer chatbots appeared in 2022. Today’s AI models can do far more than answer simple questions: they can analyze documents, research complex topics, brainstorm ideas, help with writing, and work with large amounts of context.

But getting useful results isn’t simply about having access to a powerful AI model. It’s about knowing how to work with it.

That’s where AI prompting comes in.

In this course, Andrew Ng explores the practical techniques that can help you move from using AI for simple questions to using it as a powerful partner for research, problem-solving, brainstorming, writing, and more.

From AI Novice to AI Power User

One of the central ideas in the course is the difference between an AI novice and an AI power user.

A novice might treat AI much like a search engine: type in a short question and expect an answer. That can work for simple tasks, but more complex problems require a different approach.

An AI power user gives the model enough information to understand the task.

For example, instead of simply asking AI to write a self-review, you could provide information about your projects, accomplishments, notes, and other relevant material. With that additional context, the model has much more to work with and can produce a response that is specific rather than generic.

The key lesson is simple:

The quality of your AI output depends heavily on the quality of the information and instructions you provide.

Give AI the Context It Needs

Think of AI as an extremely capable new colleague who doesn’t know you yet.

If you assign that colleague a task without explaining the background, they may produce something reasonable—but generic. Give them the relevant documents, goals, preferences, examples, and constraints, and they can do a much better job.

The same principle applies to AI.

Context can include:

  • Background information
  • Documents and files
  • Examples of previous work
  • Your goals and preferences
  • Constraints or requirements
  • Feedback from previous attempts
  • Relevant data

Modern AI models can process remarkably large amounts of context. That means users can provide far more information than they may initially realize.

The challenge is not simply giving AI more information. It is giving it the right information.

Ask AI for Honest Feedback

Another important prompting skill is learning how to reduce the tendency of AI systems to simply tell you what you want to hear.

For example, if you tell an AI that you have a “great business idea” and then ask it to critique the idea, you have already suggested the answer you expect.

Instead, you can ask the AI to evaluate the idea objectively using specific criteria.

For example:

Analyze this business idea objectively. Evaluate the problem, market opportunity, and competitive advantage. Identify both strengths and weaknesses and explain your reasoning.

You can also provide a rubric or grading criteria.

This gives the AI a clearer framework for its response and makes it less dependent on the assumptions embedded in your question.

Don’t Ask AI to Write Everything in One Step

A common AI workflow is:

Prompt → Generate → Copy

But the course presents another approach: treat AI as a thinking partner.

Instead of immediately asking AI to write a finished article, start by asking it to create an outline. Review the outline, provide feedback, and iterate. Then develop the outline into bullet points before eventually asking AI to produce the final draft.

This creates several opportunities to improve the result before committing to the final text.

The process becomes more like collaboration:

Ideas → Outline → Feedback → Refinement → Draft → Editing

This can help avoid generic, repetitive AI-generated writing and produce content that is more closely aligned with what you actually want.

Understand Where AI Gets Its Knowledge

To use AI effectively, it also helps to understand where its knowledge comes from.

AI models learn patterns from very large amounts of text and other training data. This can include material from websites, books, encyclopedias, research articles, forums, social media, and other sources.

But this knowledge has limitations.

Some topics appear frequently across the internet, while others are much more niche. AI can also encounter outdated information or misconceptions in its training data.

This gives users an important prompting principle:

Don’t assume that everything the AI knows is current, complete, or equally reliable.

For questions that depend on current information, another capability becomes important: web search.

Know When to Use Web Search

An AI model’s pre-trained knowledge is effectively frozen at a particular point in time. Events that happen after that point may not be included in its original training.

Web search allows an AI system to supplement its existing knowledge with more recent information.

Web search can be particularly useful for:

  • Current events
  • Recent developments
  • Local information
  • Current recommendations
  • Niche topics
  • Information that may have changed recently

However, web search isn’t automatically a guarantee of accuracy.

AI can encounter outdated, inaccurate, or low-quality sources. The course therefore emphasizes the importance of being deliberate about the sources you want the AI to use.

For example, when researching a scientific or health-related question, asking the model to prioritize official organizations and rigorous research can help steer it toward more authoritative information.

Web Search vs. Deep Research

Not every research task requires the same level of effort.

The course introduces three broad approaches to finding information:

  1. Pre-trained knowledge – useful for common, relatively stable information.
  2. Web search – useful when you need current information or information from a handful of sources.
  3. Deep research – useful when a question requires extensive research and synthesis across many sources.

Deep research can involve multiple searches, evaluating sources, conducting additional searches, and eventually synthesizing the findings into a detailed report.

This makes it particularly useful for questions where you need to understand several dimensions of a complex topic rather than simply retrieve a fact.

A useful way to think about the distinction is:

Simple question → AI’s existing knowledge

Current or niche question → Web search

Complex, multi-source question → Deep research

Use AI as a Brainstorming Partner

AI can also be valuable when you don’t know exactly what you want yet.

One obvious use is generating a large number of ideas. But the course goes further: effective brainstorming involves giving AI context and then iterating on its suggestions.

For example, a generic request for a workout plan might produce predictable suggestions such as squats and push-ups.

Add unusual but relevant context—such as available equipment, time constraints, motivation challenges, or personal preferences—and the AI has more information from which to generate customized possibilities.

This leads to a useful brainstorming workflow:

Give context → Ask for several options → Give feedback → Ask for more options → Repeat

Rather than asking AI to decide everything for you, you use it to expand the range of possibilities and then apply your own judgment.

Iteration Is One of the Most Powerful Prompting Techniques

You don’t need to create the perfect prompt on your first attempt.

In fact, one of the course’s key lessons is that conversation itself can provide context.

Suppose AI gives you three possible approaches. You can tell it:

  • Which option you prefer
  • Which option you dislike
  • Why you feel that way
  • What information you forgot to provide
  • What constraints have changed

The AI can then use that feedback to generate another set of options.

This creates a feedback loop:

Generate → Evaluate → Give feedback → Generate again

The more useful feedback you provide, the better the AI can understand what you are looking for.

Manage Your AI Context

AI models can work with much more information than a typical human can hold in working memory.

That makes context powerful—but it also means context needs to be managed.

When you continue a conversation, the model has access to the previous messages and information you’ve provided. This can be helpful when you’re iterating on the same task.

But if you suddenly switch to an unrelated task, the previous context may become irrelevant or distracting.

For example, if you’ve been discussing your own workout preferences and then suddenly ask AI to create a workout plan for someone else, much of the existing context may no longer be useful.

In those situations, starting a new conversation can help give the model a cleaner context for the new task.

The Bigger Opportunity: AI as a Work Partner

The most important shift isn’t learning a collection of clever prompts.

It’s changing how you think about AI.

Instead of treating AI as a search box that produces answers, you can use it as a thinking partner—something that helps you explore ideas, analyze information, research topics, develop drafts, and iterate on solutions.

The course demonstrates how this approach can extend beyond chat. AI tools can work with files, search the web, perform research, and, in some applications, interact with files on your computer to help complete real tasks.

That changes the role of prompting.

You’re no longer just asking:

“What is the answer?”

You’re increasingly asking:

“How can I work with AI to solve this problem?”

Start Building Your AI Prompting Skills

AI prompting is becoming less about memorizing the perfect wording and more about developing good working habits:

  • Give AI relevant context.
  • Be clear about your desired outcome.
  • Ask neutral questions when you want objective analysis.
  • Use rubrics and criteria when appropriate.
  • Iterate instead of expecting perfection from the first response.
  • Use web search when information needs to be current.
  • Use deep research when a question requires extensive source synthesis.
  • Give AI feedback on its ideas.
  • Manage the context you provide.
  • Treat AI as a collaborator and thinking partner rather than simply an answer machine.

These techniques can help you get substantially more value from the AI tools you already use—and provide a foundation for exploring more advanced AI workflows.

Watch the Full AI Prompting Course

Want to see these techniques demonstrated step by step, with practical examples and hands-on exercises?

Watch the full video course and follow along with the examples to start developing your AI prompting skills today.

Watch the video

Prefer to learn by watching?

Open the dedicated video page to watch this lesson.

Watch video →

Share with