This free AI course is a collection of short lessons, each built around one real work task, that you can read and practise today with no sign-up. Every lesson shows a messy starting point, a step-by-step method, a weak and an improved result, and how to check the output before anyone else sees it.
By the end of this lesson you can use AI for presentations the reliable way: decide the audience and the one message, have the AI draft a storyline and speaker notes from your material, and check every number yourself before you build a single slide. The method works whether you use ChatGPT for PowerPoint decks, Claude, Gemini or Copilot.
Start the lessonThis lesson shows you how to turn raw notes or a transcript into AI meeting notes that people act on: decisions, owners, deadlines and open questions, each traceable to what was actually said. You will also learn when you need consent to record, and how to check the summary before you send it.
Start the lessonThis lesson teaches AI blog writing in four stages, brief, outline, draft and edit, so the post sounds like your company and every claim in it is one you can stand behind. It works for any AI for writing tool your team is allowed to use.
Start the lessonThis lesson shows you how to use AI report writing for recurring status reports: you supply the notes and the figures, the AI gives them a consistent structure, and you check and own every number. Once the template works, each report takes a fraction of the drafting effort and reads the same way every month.
Start the lessonYou do not need a dedicated app: with the right prompt, any AI assistant works as an AI SOP generator that turns a recorded walkthrough or rough notes into a standard operating procedure with steps, owners, checks and exceptions. This lesson shows the method, a worked example and how to test the procedure before anyone relies on it.
Start the lessonThis lesson shows managers how to use a ChatGPT performance review prompt, or the same method in Claude, Gemini or Copilot, to structure evidence they have already gathered into balanced, specific feedback. The AI never invents feedback, never sees more personal data than your employer's approved tool allows, and never makes the rating decision.
Start the lessonThis lesson shows you how to write an AI job description that describes the real job, not a generic one, and then use the same tool to find wording that could put good candidates off. The method works as a ChatGPT job description prompt or in Claude, Gemini or Copilot, and it ends with HR checking anything legal.
Start the lessonThis lesson teaches a reliable way to use AI for social media posts: start from one real source idea, have the AI adapt it for each platform in your brand voice, and publish only claims you can prove. Whether you use ChatGPT for social media or another assistant, the method is the same.
Start the lessonCorrelation is not causation means that two numbers moving together does not prove that one makes the other happen. In this lesson you will learn the three usual explanations for a misleading pattern, confounders, reverse causation and chance, and how a simple experiment tells you what is really going on.
Start the lessonNeural networks are models built from layers of simple connected units that learn patterns by repeatedly adjusting the strength of their connections to reduce their errors. This lesson explains how they work without equations, why they need so much data, where they are used and what they cannot do, so you can question a vendor or a colleague with confidence.
Start the lessonOverfitting is when a model learns its training data so closely that it fails on new data; underfitting is when it is too simple to learn the pattern even in its training data. This lesson shows you how to tell them apart by comparing training and test results, using a fictional customer churn model, and what to do about each.
Start the lessonRecommendation systems are models that predict which items a person is most likely to want, based on the items themselves, on what similar people chose, or on both. This lesson explains the two main approaches, why new products and new customers are hard, and how recommendations can quietly narrow what people see.
Start the lessonGood dashboards and KPIs show a handful of measures, each tied to a decision someone will make, rather than every number that can be tracked. This lesson shows you how to choose those measures, mix leading and lagging indicators, lay out a dashboard people actually read, and use AI to draft commentary without letting it invent explanations.
Start the lessonA small set of Google Sheets formulas, lookups, FILTER, QUERY, ARRAYFORMULA and a few text functions, replaces most of the copying and pasting people do in spreadsheets each week. This lesson explains what each one does in plain words, when to use it, and how to get Gemini or another AI assistant to draft a formula you then check.
Start the lessonData cleaning with pandas means loading a messy file into a DataFrame and fixing missing values, duplicates, untidy text, numbers stored as text and inconsistent dates, in a repeatable order. This lesson explains each step in plain words and names the pandas functions involved, so you know what to write or what to ask an AI assistant to write.
Start the lessonEthical AI at work means asking, before you act on an AI output, whether it is fair, open about how it was made, respectful of people's data, owned by a named person and unlikely to cause harm. This lesson turns those five ideas into questions you can answer in minutes, using everyday office decisions.
Start the lessonResponsible AI is the set of practices that make good intentions about AI happen every time: a clear policy, a review step, simple documentation, ongoing monitoring and a way to handle mistakes. This lesson shows a small team how to set those controls up for one real AI use, without a compliance department.
Start the lessonHuman in the loop means a person reviews AI work at a point where they can still change the outcome, and has the time, information and authority to do it. This lesson shows you where that point should be, which review design fits which task, and how to stop a review step from turning into a click-to-approve habit.
Start the lessonEvaluating sources means checking who is behind a claim, how current it is, what evidence supports it and what bias might shape it, then confirming it elsewhere before you use it. This lesson applies those checks to a statistic an AI assistant handed you, because AI citations need the same scrutiny as any web page, and sometimes more.
Start the lessonGenerative AI works by learning patterns from very large amounts of data and then producing new text or images that fit those patterns, one small step at a time. Once you see that a chatbot predicts likely next pieces of text rather than looking up facts, its strengths, its varied answers and its invented details all make sense.
Start the lessonRetrieval-augmented generation, usually shortened to RAG, is a way of building an AI assistant that first finds the relevant passages in a set of documents and then writes its answer from those passages. This lesson explains how it works, when it is the right approach and how it fails, without needing any code.
Start the lessonStructured output means asking an AI model to return data in a fixed shape, usually JSON that follows a schema you define, so your app can read it without guessing. This lesson explains why that matters, how the major model APIs support it, and how to validate results and recover when they fail, all in prose with no code listings.
Start the lessonIncident response is the set of steps you take from the moment something looks wrong until normal work resumes: recognise, contain, report, recover and learn. This lesson walks a small team through the first hour using official UK NCSC and US CISA guidance. It is a basic awareness lesson for staff and managers, not professional incident-response training.
Start the lessonMalware is any software designed to harm a device or steal from it, and ransomware is the kind that locks or steals your files and demands payment. This lesson covers the main types, the warning signs and the three habits that do most of the protecting: updates, offline backups and multi-factor authentication. It draws on NCSC and CISA guidance.
Start the lessonPhishing and social engineering work by pushing you to act before you think, so the defence is a habit: pause, verify through a channel you already trust, then report. This lesson covers email, text, voice and AI-written lures and the pressure tactics they share, using NCSC and CISA guidance.
Start the lessonAI agent security comes down to one rule for everyday users: an agent should only be able to do what the task needs, and anything risky should wait for your approval. This lesson explains prompt injection, over-broad permissions and data leakage in plain terms, using the OWASP Top 10 for LLM Applications, and shows how approvals and logs keep you in control.
Start the lessonA smart contract is a program stored on a blockchain that runs exactly as written when someone sends it the right transaction, with the result visible to everyone. This lesson explains how smart contracts work, what they are used for and where they fall short, using ethereum.org's documentation. It is an educational lesson, not investment or legal advice.
Start the lessonSmart contract security is the practice of finding flaws before deployment, because a deployed contract usually cannot be patched and anyone can call it. This Code lesson explains the main vulnerability classes in prose, then the testing, review and audit steps that catch them, using ethereum.org and the OWASP Smart Contract Top 10. It is educational and not investment advice.
Start the lessonA process map shows who does what, in what order, where work changes hands and where a decision sends it down a different route. This lesson shows you how to draft process maps with AI from rough notes, and then check the draft with the people who actually do the work, because a map nobody recognises is worse than no map.
Start the lessonGood release notes tell each reader what changed for them, what they need to do and when, in words they understand. This lesson shows you how to draft release notes from tickets with AI for customers, internal teams and developers, and how to check that every line describes what actually shipped.
Start the lessonUseful competitor research answers a specific decision with a comparison where every cell points to evidence a colleague can open and check. This lesson shows you how to use AI to frame the questions, read official sources and fill the table, without letting it fill gaps with confident guesses.
Start the lessonEach lesson teaches one method for one task, such as turning meeting notes into decisions and owners, writing a job description without exclusionary wording, or building a presentation outline before a single slide exists. You get the situation, the method, a worked example with a fictional company, a checking routine, the common mistakes, a practice brief with a rubric and a few copyable prompts.
The lessons are tool-neutral. The methods work in ChatGPT, Claude, Gemini or Copilot, because they depend on what you give the tool and how you check what comes back, not on a particular feature. Most people finish a lesson in 10 to 15 minutes and can use the method at work the same day.
Yes, the lessons on this page are free AI training: free to read, with no account, no email and no time limit. You can practise the task on your own in whichever approved AI tool your employer allows.
What the free lessons cannot do is look at your own work. A Tosheo.ai plan adds a short diagnostic, a personalised path of 10 to 15 minute lessons matched to your role and level, hands-on tasks marked against a visible rubric with specific feedback on what you actually wrote, and a final practical assessment that leads to a recognised certificate. The certificate comes only with a plan; it is not part of the free lessons and it is not an accredited qualification. Every plan starts with a 7-day free trial, and no credit card is needed.
The full text of every lesson, the worked examples, the practice brief, the rubric and the prompts.
A diagnostic, an adaptive path for your role, rubric feedback on your own work, progress tracking and the recognised certificate.
Start with the task you did last week and will do again next week. If you run meetings, begin with meeting notes. If you report to a manager or a client, take the report writing lesson. Marketers usually start with blog writing or social media posts, HR teams with performance reviews or job descriptions, and operations teams with SOP writing. If you want ready-made prompts rather than a full lesson, the prompt library collects checked prompts for common tasks.
Yes. Every lesson on the Tosheo.ai lessons page is free to read, with no account, no email address and no time limit. Paid plans add personalised paths, rubric feedback on your own work and a recognised certificate.
No. The free lessons do not include a certificate. The recognised certificate comes with a Tosheo.ai plan after a final practical assessment, and every plan starts with a 7-day free trial with no credit card needed.
Any general AI assistant your employer approves, such as ChatGPT, Claude, Gemini or Copilot. The methods depend on the inputs you give and the checks you run, so they work across tools.
Most lessons take 10 to 15 minutes to read and practise once. Repeating the practice brief with your own real task is where the method sticks.
Yes. The lessons assume you can open an AI assistant and type a request, nothing more. Each one explains the method step by step and shows a weak and an improved result.
Explore role-specific lessons, practical tasks and the manager view.
