AI SKILL LAB
LABRUStart application
GUIDE · AI SKILL LAB

Where should adults start with AI? First practical tasks

You do not have to begin learning AI by exploring dozens of services, complex prompts or automation. A practical first step is to take one small task from your own work, produce a draft result and check it yourself. AI Skill Lab builds its adult learning path around a real task: set a goal, define quality criteria, build a solution and verify the result. S01 S02

Checked 8 October 2026 · next review 8 January 2027

Your first task: a small result you can verify

Choose a task for which you already have source material and a clear criterion for completion. For example, turn anonymised notes into a plan, draft an email or compare options using criteria you have defined. These are practice exercises, not a promise that they will be effective in every case.

Before your first prompt, ask yourself four questions:

  1. What do you need to produce? For example, an action plan, an email, a list of questions or a comparison table.
  2. What should the answer be based on? Your verified points, public text or anonymised notes.
  3. What does a good result look like? Specify the length, format, audience and quality criteria.
  4. Who will check the answer? You will — especially dates, names, links, numbers and decisions with consequences.

In AI Skill Lab's adult programme, this approach means working from a task towards a verifiable outcome; learning tools does not replace quality criteria. S01 S02

Four options for your first exercise

TaskWhat to give the AIWhat to produceWhat to check yourself
Plan from notesA few anonymised pointsA sequence of steps with open questionsWhether the AI invented tasks, deadlines or commitments
Email draftThe purpose of the message, the recipient's role and the required toneA short, editable messageFacts, tone, promises and addresses before sending
Compare optionsYour options and decision criteriaA table of advantages, disadvantages and unknownsWhether facts are distinguished from assumptions; whether enough information is available
Work with public materialA link or text that you are allowed to useA brief outline and questions to verifyWhether the conclusions match the source; whether reliable sources are available

Start with one option. Check outside information, especially recent information, against the primary source: a coherent explanation does not prove that every fact is correct. AI Skill Lab's learning path explicitly includes sources, checks for errors, edge cases and uncertainty. S02

One prompt with a clear structure

OpenAI recommends clearly specifying the task, context and desired format, and refining your prompt through conversation when necessary. You do not need a 'secret' set of commands: specific requirements are more useful than a vague request. S03 S04

Practice template — five prompt fields:

  1. Task: help me with [one specific task].
  2. Context: [only the necessary anonymised or publicly available information].
  3. Facts: treat only the information I provided as facts; if information is missing, list questions rather than making guesses.
  4. Result: [table / list / short draft] for [audience], length [limit].
  5. Verification: show separately what needs to be checked manually and what information remains unknown.

Example: you need to organise notes before a work meeting, but you do not want to share names, phone numbers or confidential details. Replace them with neutral labels such as 'task A' and 'question B'. Ask for a list of topics and questions. This is a hypothetical exercise; it does not describe actual client results.

Checking the answer: the work does not end in the chat

Compare the answer with your original notes. Check that the meaning has not changed and that the AI has not introduced dates, links, numbers or commitments on behalf of others that you did not specify. OpenAI separately warns that ChatGPT may produce incorrect facts and invented links or quotations; important information should be checked against reliable sources. S09

A simple process for practice is:

  1. Identify all specific claims that could be incorrect.
  2. Check facts against the provided material or an available primary source.
  3. Mark assumptions and information that has not been established.
  4. Correct the text yourself, then use it in external communication.

If a task involves health, law, finance, safety or other decisions where errors can have serious consequences, do not treat an AI answer as a standalone professional opinion. Consult a relevant specialist if needed. OpenAI also recommends checking important information even when an answer sounds convincing. S09 Fact-checking and human checkpoints are part of AI Skill Lab's approach. S02

A second pass: refine what did not work

Ask the model to change a specific part of the answer: its structure, length, tone or decision criterion. Gradual refinement is one of the techniques recommended by OpenAI. S03

Example refinement: 'Keep the verified points. Remove repetition, shorten the plan to five actions and mark everything that needs clarification. Do not add new facts.'

After the second pass, compare the result with the source material again. The number of iterations does not, by itself, demonstrate quality.

Data, privacy and boundaries of authority

Do not enter passwords, API keys or other secrets in practice prompts, even when the processing of ordinary data is permitted. OpenAI recommends not sharing API keys or placing them in client applications and repositories. S10 Do not share financial account details, confidential documents or other people's personal data without appropriate permission and an understanding of the processing terms. Public or anonymised examples are enough for initial exercises.

If you use ChatGPT, check Data Controls separately: OpenAI describes settings that let you control the use of conversations to improve models, including Improve the model for everyone. The settings available depend on your account; turning off training does not replace the rules for handling confidential data. S05

Start with a draft, not an autonomous action. Sending emails, changing files, publishing and accessing other services are separate actions: it is important to define permissions, confirmation and a way to verify the result in advance. This principle is consistent with AI Skill Lab's published rules for controlling an agent's external effects. S01

What to keep after your first exercise

Keep three things:

  • your original task and a safe set of inputs;
  • the prompt that produced the most suitable draft;
  • the corrected result and a short verification checklist.

This will help you repeat the exercise and explain why the result is suitable for you. AI Skill Lab's learning programme includes reusable work templates, quality assessment and handover of a working process to the learner. S02

Once one task is clear and repeatable, you can move on to working with documents, research, automation or assistants. You do not need to do everything at once. For more about the practical format, see personal learning and the curriculum. S01 S02

Sources

Public sources checked: 8 October 2026. The links below support the methodological statements in this draft. They do not replace final fact-checking and editorial review.

How to reach us

Use the application form or a direct channel; for a minor, coordination stays with an adult.

Start application
TelegramWhatsApp+66 64 970 1204LINEiwf555Emailrobert@aiskillab.work

Reply within 1–2 business days.