Learning AI · Practical guide
Learn applied AI by building something you can check.
Start with a small problem you understand. Work with examples and build a solution whose results you can verify. Learning to describe the task and detect errors matters just as much as learning to use a tool.
01Decide what you want to be able to do
Separate three goals: using AI in your own work, building a solution for your business and delivering projects to other businesses. They share foundations, but each requires different practice.
If you want to build applications, practise defining screens, data and actions. If you want to automate tasks, focus on inputs, rules, connections and exceptions. A small project gives your learning a clear direction.
02Learn to describe inputs and expected results
A useful instruction explains the task, supplies context and defines how the result should be presented. Add examples when they help distinguish a correct answer from one that only sounds convincing.
Practise checking the response as well. Separate what you can verify with data from what needs judgement or human review.
03Try a first exercise with fictional data
Write ten fictional customer requests and define three categories. Ask an AI tool to classify each request and identify any missing information. Compare its answers with classifications you prepared yourself.
Add one ambiguous request and one incomplete request. Decide when the workflow should ask for help. Once you understand its behaviour, consider connecting it to a form or another business tool.
- Write down what the workflow receives and what it should return.
- Prepare successful examples and cases that should stop the process.
- Record errors and change one instruction at a time.
- Repeat the checks before connecting real systems or data.
Reference [1]: GOV.UK Service Manual
04Understand the difference between building and delivering
A solution you deliver needs instructions, access controls and a way to resolve errors. If other people depend on it, decide who maintains it and what happens when a connected tool changes.
If you offer AI services, describe the problem you solve, the scope and how the work will be accepted. A precise offer is more useful than promising to automate anything.
05Look for practice and useful feedback
When comparing AI training, check whether you will work on projects, receive feedback and learn to assess what you build. Ask about prerequisites and the current programme before making a decision.
ELITE IA is my Spanish-language training programme and community for learning to build, sell and deliver AI solutions. Its existing website explains the approach and the programmes available.
Sources and further reading
These references expand on the concepts indicated. The examples and exercises are original editorial material.
[1] GOV.UK Service Manual
How the alpha phase works ↗
Further reading on testing ideas through prototypes.
Back to the related section
Frequently asked questions
What should I learn about AI first?
Start by defining tasks, supplying context and checking responses against examples. Then choose a small project that makes you work with inputs, outputs and errors.
Can I learn AI without a technical background?
Yes, you can start with exercises and assisted tools. Building solutions for real use means gradually learning about data, integrations, access controls and testing.
Keep reading
Learning AI
How to check an AI assistant’s answers before using it ↗
Build a simple test with questions, evidence and acceptance criteria. Use a worked exercise to find unsupported answers before relying on an assistant.
Learning AI
Turn meeting notes into tasks with AI: a checkable exercise ↗
Practise extracting commitments, owners and deadlines without inventing them. Includes fictional notes, a reusable instruction and a reference answer.
Learning AI
How to test an app idea without building the full backend ↗
Create a prototype with fictional data to check a user journey. Separate what the interface demonstrates from what still needs implementation.