How to Learn AI from Scratch in 2026: A Practical 3-Level Roadmap

Artificial intelligence is changing how people research, write, analyze information, and solve problems. But learning AI can feel overwhelming when new tools and techniques appear every week. Instead of trying to learn everything, focus on three practical skills: mastering one AI assistant, providing better context, and building reusable workflows.

Level 1: Choose One AI Assistant and Learn It Well

Start with one assistant and learn how to use it consistently. Three popular choices are ChatGPT, Claude, and Google Gemini.

ChatGPT: General-purpose assistance

ChatGPT can help with research, writing, brainstorming, coding, data analysis, and everyday problem-solving.

Claude: Writing and coding

Claude is worth exploring for drafting documents, editing content, working with code, and analyzing complex information.

Gemini: Multimodal tasks

Gemini is useful for working with text, images, audio, and video, particularly if you use Google Workspace. These are general starting points, not permanent rankings. The best option depends on your tasks, available features, budget, and preferences.

Should you pay for AI?

Start by evaluating the tools you already have access to. If you have access to a paid subscription through work, test whether its additional capabilities help you complete tasks more effectively. For important work, select a model suited to the complexity of the task rather than automatically accepting the default. Your first action: Choose one assistant and use it for three real tasks this week: researching a topic, summarizing a document, and improving a piece of writing.

Level 2: Focus on Context, Not Complicated Prompts

Writing long prompts is not always the best way to get better results. A clear objective combined with relevant context can make a significant difference. Remember this simple framework: OC = Outcome + Context

• Outcome: What do you want the AI to accomplish?

• Context: What information does it need to do the job well?

For example, instead of asking an AI to write a professional report, provide your notes, explain the report's purpose, and share an example of a previous report you liked. Then ask: "Create this week's report using my notes. Follow the structure and tone of the example, and highlight the main achievements, risks, and next steps." The AI now has both a clear objective and an example of what good work looks like.

Three ways to provide better context

1. Use established frameworks Frameworks provide ready-made approaches to solving problems. For example, the Pyramid Principle can help structure business communication around a central conclusion supported by logical arguments. You can ask an AI assistant to recommend an appropriate framework for your task. 2. Provide real examples Previous reports, approved emails, and successful documents can communicate your preferred tone, formatting, and level of detail. Share relevant examples and ask the AI to follow their patterns when producing new work. 3. Connect relevant tools Where supported, integrations can let AI assistants retrieve information from tools such as Google Drive, Gmail, Slack, or Notion. Check permissions and avoid sharing confidential information with unapproved services.

Create Reusable AI Projects

If you repeatedly perform the same type of work, save your instructions and reference materials instead of explaining everything from scratch each time. Depending on the platform, you can use ChatGPT Projects, Claude Projects, or Gemini Gems. A useful workspace can contain three elements:

• Instructions: Your goals, preferences, rules, and constraints.

• Knowledge files: Reference documents, templates, examples, and frameworks.

• Relevant history or memory: Important decisions and updates, where supported.

For example, create a job application workspace containing your CV, target roles, preferred writing style, and examples of strong cover letters. When a new opportunity appears, the AI can use that context to help tailor your application. Markdown files (.md) are convenient for text-based reference material because they provide a simple, structured format. PDFs remain useful when layout and visual information matter.

Level 3: Connect Individual Workflows into an AI System

Separate AI projects may not automatically share information. A broader AI system can coordinate information and workflows across multiple tasks. Imagine maintaining three projects:

• A career development plan.

• A collection of learning resources.

• A weekly productivity plan.

A connected workflow could compare these projects, identify overlapping goals, and suggest learning activities that fit your schedule. This helps uncover connections that may be missed when each project is managed independently.

Turn feedback into reusable rules

Suppose an AI assistant drafts a LinkedIn post. You edit it to make the wording more natural, remove unnecessary phrases, and improve the structure. Ask the assistant to compare the draft with your final version and identify your recurring preferences. You can then add those preferences to your project instructions for future drafts. Not every AI tool automatically learns permanently from corrections. Sometimes you need to update the instructions or workflow yourself. The principle is to turn repeated feedback into reusable guidance.

Start simple

You do not need complex automation or coding tools to begin. Start with the features available in your preferred assistant. As your workflows become more sophisticated, explore integrations and customizable tools such as Claude Code or OpenAI Codex if they suit your needs. Focus on solving real problems rather than building unnecessary complexity.

A Practical 30-Day Plan for Learning AI

Week 1: Choose your tool

• Select ChatGPT, Claude, or Gemini.

• Identify three repetitive tasks.

• Test the assistant on real examples.

• Compare its output with your usual approach.

Week 2: Improve your context

• Collect examples of successful work.

• Create reusable instructions.

• Practice the Outcome + Context framework.

• Experiment with reference documents.

Week 3: Create a dedicated project

• Set up a workspace for one recurring activity.

• Add instructions, reference files, and examples.

• Use it for several related tasks.

• Refine the instructions when you notice repeated mistakes.

Week 4: Build a repeatable workflow

• Identify the steps you repeat most often.

• Create a checklist or reusable process.

• Connect approved tools where appropriate.

• Review results and identify tasks that still require human judgment.

At the end of the month, evaluate whether AI is saving time, improving quality, or helping you accomplish more.

Common Mistakes to Avoid

Switching tools too frequently. Experiment when useful, but spend enough time with one assistant to understand its capabilities. Writing complicated prompts without useful context. More words do not automatically produce better results. Trusting every answer. AI can make mistakes or provide outdated information. Verify important claims. Repeating the same instructions. Save useful guidance and examples in reusable workspaces. Automating too early. First establish a process that works. Automate it when the benefits outweigh the complexity.

Final Thoughts: Learn AI by Using It

Learning AI does not require mastering every new model or following every trend. Start with one assistant. Learn to provide meaningful context. Save what works in reusable projects. Then connect your workflows when you have a clear reason to do so. The real advantage comes from consistently applying AI to meaningful tasks, evaluating its results, and improving your process over time. Your next step: Choose one repetitive task you perform every week and use AI to improve it today. Save the instructions that work, review the result, and build from there. This article is an independently written educational guide inspired by a YouTube video about learning AI through model selection, context engineering, reusable projects, and connected workflows. Original creator: To be added. Original video title: To be added. Original video URL: To be added before publication. Visit the original video for the creator's complete explanation and demonstrations.

Original Video and Creator Attribution

This article is an independently written educational guide based on ideas discussed in a video by Jeff Su. It includes additional explanations and practical takeaways for Hirely readers. This article is not affiliated with or endorsed by the original creator.

Original creator: Jeff Su

Original video: How I'd Learn AI From Scratch in 2026 (skip the useless 80%)

Watch the original video for Jeff Su's complete explanation and demonstrations.