How do I get started with AI?

TrainingUpdated Sep 2026

Free AI courses to get you started.

This page is for getting started with AI, whichever way you learn. Some courses build your basic understanding of what AI is and does. Others put you in front of a tool and have you build something by the end of the hour. All of them are free to take.

Claymation scene: a robot in a graduation cap raises its hand at a school desk while a businessman writes FREE on the blackboard. A sign reads: new lesson.

Every course, at a glance

One line on what each course does for you. Tap a name for the full card: how it is taught, what you build, and where it costs.

Start with the basics

  • Elements of AI

    The University of Helsinki’s introduction to what AI is and is not, with quizzes and no code.

    FreeBeginner

  • Learn Prompting

    A reference library of prompting techniques and a seven-day course on where prompts break.

    Free to startBeginner to intermediate

  • AI Fluency: Framework & Foundations

    Anthropic’s short course on working with AI well: what to hand over, how to ask, how to check the answer.

    FreeBeginnerAbout 4 hours

Courses where you build something

  • AI Summer Adventure

    Twenty-three short build-it-yourself stations, from everyday assistants to agents that run on their own.

    FreeBeginner to intermediate

  • OpenAI Academy

    OpenAI’s free home for learning ChatGPT at work: three short courses with certificates, plus live sessions and walkthroughs.

    FreeBeginner to intermediate60 to 90 minutes per course

  • AI Prompting for Everyone

    Andrew Ng’s no-code course on asking AI for what you want, with browser exercises.

    Free to startBeginnerAbout 7 hours

Courses about AI agents

  • Agents and Workflows

    OpenAI’s 90-minute course on handing an agent a piece of work: set the context, the output and the limits, then check the result.

    FreeBeginner to intermediateAbout 90 minutes

  • Building Effective Human-Agent Teams

    Anthropic’s 45-minute course for team leads on moving from one person using AI to a team working with agents.

    FreeBeginner45 minutes

  • AI Agents Course

    The free course most people name first for agents: four units from what an agent is to one you build and enter on a leaderboard.

    FreeIntermediateAbout 15 to 20 hours

  • AI Agents for Beginners

    Microsoft’s open eighteen-lesson course on building agents in Python, with a video and runnable code for each lesson.

    FreeIntermediate to advanced

  • 5-Day AI Agents Intensive

    Google’s five-day agents course, now self-paced: a short paper, a podcast and two hands-on labs a day.

    FreeIntermediateFive days, at your own pace

  • Agentic AI

    Andrew Ng’s course on the four patterns behind every agent, built from scratch in Python before any framework.

    Free to startIntermediate to advanced

Built for your department

  • AI for Finance and Accounting

    Twelve lessons that put AI to work on statements, board reports and forecasts.

    FreeBeginner to intermediate

  • Free AI HR Course

    A one-hour sprint that leaves you with a job-ad prompt, a screening workflow and a prompt library.

    FreeBeginner60 minutes

  • AI for Legal Professionals

    Contract triage, research memos and the rules on confidentiality, built for lawyers.

    FreeBeginner to intermediate

  • AI for Marketing

    HubSpot’s six lessons on taking one brief through a campaign and reading the results.

    FreeBeginner to intermediate2 hours 49 minutes

For people who code

  • AI for Beginners

    Microsoft’s twelve-week curriculum on how neural networks work, in Python notebooks.

    FreeIntermediate to advancedTwelve weeks

  • Practical Deep Learning for Coders

    fast.ai’s hands-on course for people who code: train and deploy real models from lesson one.

    FreeAdvanced

Start with the basics

What AI is, what it does well, and how to ask it for things. No code, and nothing to install. Take one of these if you want to understand the tool before you lean on it.

  • FreeBeginner
    Elements of AI, as it looks on screen.

    Elements of AI

    University of Helsinki and MinnaLearn

    Interactive quizzes and logic problems — no maths, no code, no videos. Helsinki's own count: two million students. It predates chatbots, so it teaches the foundations rather than ChatGPT.

    What you build

    • What the technology is actually doingHow a prediction gets made, and where machine learning differs from ordinary software that follows rules.
    • Judgement about its limitsBias in training data, and the failures that are built into the method rather than fixable by a better vendor.

    “Conceptually rigorous without being technical” Hashmeta review · May 2026

  • Free to startBeginner to intermediate
    Learn Prompting, as it looks on screen.

    Learn Prompting

    Learn Prompting

    A reference library of prompting techniques, plus a seven-day course on where prompts break. It was written before ChatGPT existed; its own count since is three million readers.

    What you build

    • A method you can reuseThe named techniques, organised, so you can look one up rather than remember it.
    • Knowing how prompts failThe security material on prompt injection is unusual and worth the time if anything you build faces outward.

    “one of the largest online resources for learning prompting” Louis Bouchard · Apr 2023

    Where it costsThe reference docs and the seven-day course are free. The certification tracks are paid.

  • FreeBeginnerAbout 4 hours
    AI Fluency: Framework & Foundations, as it looks on screen.

    AI Fluency: Framework & Foundations

    Anthropic, with Rick Dakan (Ringling College) and Joseph Feller (University College Cork)

    Fourteen short video lessons and a quiz around one framework with four parts: delegation, description, discernment and diligence. About four hours, no code, a free Claude Academy login, and a certificate at the end.

    What you build

    • A rule for what to hand to AIDeciding which parts of a task the tool should do and which stay with you, before you open it.
    • Requests that get useful answersDescribing the job, the audience and the format so the first draft is close.
    • A habit of checkingReading an answer for what is wrong or missing, and being open about where AI was used.

    “gives more context than most paid introductions” AI Weekly, the best free AI courses of 2026

Courses where you build something

Start here if you learn by doing. Each one leaves you with something built in a tool you already have, often within the hour.

  • FreeBeginner to intermediate
    AI Summer Adventure, as it looks on screen.

    AI Summer Adventure

    Nufar Gaspar, with The AI Daily Brief and Superintelligent

    Twenty-three hands-on stations in three tiers, at your own pace — from the tools you already own to workflows that run on their own. Free with an account.

    What you build

    • Everyday productivity assistantsTier one works on the daily grind — turning messy notes and transcripts into something you would actually send.
    • Custom knowledge assistantsTier two builds assistants that answer from documents you supply, rather than from whatever the model happens to know.
    • Autonomous agents and multi-tool workflowsTier three chains steps together so a task runs without you sitting over it.
  • FreeBeginner to intermediate60 to 90 minutes per course
    OpenAI Academy, as it looks on screen.

    OpenAI Academy

    OpenAI

    Three self-paced courses of 60 to 90 minutes each (AI Foundations, Applied AI Foundations, Agents and Workflows) with progress tracking and a completion certificate through a free ChatGPT login, plus live sessions and a library of short walkthroughs.

    What you build

    • One everyday task done betterThe first course takes a task you already do in ChatGPT and makes the output something a process can rely on.
    • A repeatable workflow planThe second course turns a routine job into steps an assistant handles in sequence, written down so a colleague could run it.
    • An agent-assisted workflowThe third course hands a bounded piece of work to an agent and keeps the review with you; it is listed separately under AI agents below.

    “practical and hands-on, and most importantly, there are no ads or paywall” Fabio Dantas, Class Central · Dec 2025

  • Free to startBeginnerAbout 7 hours
    AI Prompting for Everyone, as it looks on screen.

    AI Prompting for Everyone

    DeepLearning.AI, taught by Andrew Ng

    Twenty-one video lessons with guided browser exercises. No coding at any point — built, in its own words, for anyone who uses ChatGPT, Claude or Gemini at work.

    What you build

    • Dense reports into executive summariesWorking a long document down to the version a board would read, and knowing what got lost.
    • The model as a criticInstructing it to argue against you as a sceptical director or a strict editor, which is where it earns its keep.
    • Thinking through an ambiguous problemUsing it on scenarios that have no clean answer, rather than on tasks you already know how to do.

    Where it costsThe videos are free. Graded assignments and the certificate need a Pro membership at $25 a month, billed annually.

Courses about AI agents

An agent is software you hand a whole job to. These courses show how agents work and walk you through building one. Most expect a little Python; if you do not code, take the third tier of AI Summer Adventure above first. Read our agents page if the word is new to you. What an AI agent is, in plain words.

  • FreeBeginner to intermediateAbout 90 minutes
    Agents and Workflows, as it looks on screen.

    Agents and Workflows

    OpenAI

    Self-paced, about 90 minutes, video and short exercises inside ChatGPT, with progress saved so you can stop and return. Needs a free ChatGPT login and gives a completion certificate. The third of OpenAI Academy’s three courses; start here if you already use ChatGPT every day.

    What you build

    • A brief an agent can act onGiving the context, the output you want and the boundaries before the agent starts, so the result needs one review rather than five.
    • A review step you keepWhere a person must still look, which the course treats as part of the workflow rather than an afterthought.
    • A workflow you run againTurning the one task that worked into a pattern you can reuse and hand to a colleague.

    “useful because it connects AI literacy to repeatable work and human oversight” Curtis Pyke, Kingy AI · Jun 2026

  • FreeBeginner45 minutes
    Building Effective Human-Agent Teams, as it looks on screen.

    Building Effective Human-Agent Teams

    Anthropic

    Five short lessons and a quiz, self-paced, 45 minutes, no code. A free Claude Academy login saves your progress and gives a completion badge. Anthropic marks it beta, so expect it to change.

    What you build

    • The difference between an agent and a chatbotIn words you can repeat to your team: what an agent does on its own, and what it still hands back to a person.
    • Four working rules for a mixed teamClear roles, a written north star, a gradual release of responsibility, and the right access to information.
    • A checklist for your organisationWhat to settle before agents join the work: the norms, who owns what, and where a person signs off.
  • FreeIntermediateAbout 15 to 20 hours
    AI Agents Course, as it looks on screen.

    AI Agents Course

    Hugging Face

    Four units of about a week each, three to four hours a week, self-paced with no deadline. Reading, short videos and notebooks that run in the browser. It expects basic Python and a free Hugging Face account.

    What you build

    • A clear picture of how an agent thinksThe loop of thought, action and observation that every agent runs, and what a tool is to it.
    • A first agent in three toolkitsThe same simple agent built with smolagents, LlamaIndex and LangGraph, so you can see what a framework adds.
    • An agent scored against a benchmarkThe final assignment: an agent you build, test and submit to a public leaderboard, with a free certificate.
  • FreeIntermediate to advanced
    AI Agents for Beginners, as it looks on screen.

    AI Agents for Beginners

    Microsoft

    Eighteen written lessons, each with a short video and a Python notebook, plus a Discord for questions. Free and open source. You need Python, and the code samples expect a Microsoft Foundry account or an API key for another model provider.

    What you build

    • A working agent with tools and memoryAn agent that calls tools, keeps track of what it has done, and looks things up in your own documents.
    • Agents that work togetherSplitting a job across several agents and the patterns for planning and checking their work.
    • Something you could run for realThe later lessons cover getting an agent into production, keeping it secure, and running one on your own machine.
  • FreeIntermediateFive days, at your own pace

    5-Day AI Agents Intensive

    Google and Kaggle

    Five days of material you can take at your own pace: a technical whitepaper, a companion podcast and two browser notebooks a day, built on Gemini and Google’s Agent Development Kit. Free with a Kaggle account. The labs are in Python.

    What you build

    • Your first agent, and your first team of themDay one builds an agent that can search the web for an up-to-date answer, then a small system of several agents.
    • Agents that rememberSessions and long-term memory, and how to manage what an agent keeps in view.
    • A way to tell whether it worksEvaluation, and what it takes to move an agent from a prototype to something people rely on.
  • Free to startIntermediate to advanced
    Agentic AI, as it looks on screen.

    Agentic AI

    DeepLearning.AI, taught by Andrew Ng

    Video lessons with coding exercises. Four patterns, reflection, tool use, planning and several agents working together, each built by hand before a framework is introduced. Expects intermediate Python.

    What you build

    • An agent that checks its own workThe reflection pattern: a first draft, a critique, and a better second draft, all by the same system.
    • An agent that uses tools and plansGiving a model functions to call, then letting it break a task into steps and work through them.
    • Several agents on one jobSplitting a task between agents with different roles, and the judgement of when that helps.

    Where it costsThe videos are free. Graded assignments and the certificate need a Pro membership at $25 a month, billed annually.

Built for your department

Role-specific courses that work on your own material: your contracts, your close, your campaign brief.

  • FreeBeginner to intermediate

    AI for Finance and Accounting

    FreeAcademy.ai

    Twelve applied lessons worked directly against financial statements, reports and spreadsheets. No signup needed to start; the certificate takes a final exam.

    What you build

    • Statement analysis you can checkPulling variance drivers and ratios out of a P&L or cash flow statement, with the working visible.
    • A repeatable board-reporting templateTurning a monthly close into narrative commentary, the same way every month.
    • Pressure-testing a forecastRunning scenarios against your own assumptions and flagging expense entries that look wrong.

    “If you can use a spreadsheet and write an email, you can use these workflows” FreeAcademy.ai, from the course page

  • FreeBeginner60 minutes
    Free AI HR Course, as it looks on screen.

    Free AI HR Course

    Johannes Sundlo

    A sixty-minute sprint. You leave with four things built in your own tools, and the certificate is free — no payment details, just an email.

    What you build

    • A job-ad prompt in your voiceA role description generated the way your company actually writes, plus scoring criteria to judge candidates consistently.
    • A screening summary workflowA repeatable way to summarise applications against the role before a human decides.
    • Your first prompt libraryThe prompts you built, organised for reuse, and a one-page guardrail checklist for using AI on people decisions.

    “This is a doing course, not a watching course” Johannes Sundlo, from the course page

  • FreeBeginner to intermediate2 hours 49 minutes
    AI for Marketing, as it looks on screen.

    AI for Marketing

    HubSpot Academy

    Six lessons, fifteen videos and four quizzes, from brief through to performance analysis. Needs a free HubSpot account.

    What you build

    • One brief into a full campaignTaking a single brief out to social, email and long-form without losing the voice.
    • Testing a message before launchTrying headlines and value propositions against a described audience while it is still cheap to change them.
    • Reading the resultsTurning campaign performance data into what to change next week.

For people who code

How the technology works underneath, taught in Python. Skip these unless you write code, or want to.

  • FreeIntermediate to advancedTwelve weeks
    AI for Beginners, as it looks on screen.

    AI for Beginners

    Microsoft

    Microsoft's own pitch: twelve weeks, twenty-four lessons, with quizzes and runnable notebooks. For people comfortable in Python — it teaches how neural networks work under the hood.

    What you build

    • Working models, from scratchBuilding small networks that classify images and read handwriting, so the mechanism stops being a metaphor.
    • How language models work underneathText classification and the transformer architecture that everything else on this page is built on.
    • Bias you can measureEvaluating algorithmic bias in vision models rather than discussing it in the abstract.

    “Free, MIT-licensed, no account or certificate paywall” explainx.ai review · Aug 2026

  • FreeAdvanced
    Practical Deep Learning for Coders, as it looks on screen.

    Practical Deep Learning for Coders

    fast.ai

    Top-down coding projects in Python. fast.ai's stated prerequisite: knowing how to code, with a year of experience enough.

    What you build

    • A trained model in lesson oneFine-tuning a real vision model on your own images before any theory is explained.
    • Models tuned to your own textTaking an open-weight model and adapting it to a specialised body of documents.
    • Something deployedPackaging a trained model into an application, which is where most courses stop and this one keeps going.

    “Your courses completely changed my perspective on what was possible for me” Hacker News · Jul 2022