How to use AI for learning at work

An AI can write a convincing answer to a difficult customer. That still leaves a question: could you give the answer yourself if the customer called?

Getting help with a piece of work and learning how to do it are both useful. They need different instructions. If the deadline is in ten minutes, you may want a draft. If you want to get better at the task, you need a chance to attempt it, notice where you struggle, and try again.

Using AI for learning at work becomes more practical when you make that choice explicit. Give it a role in your practice session: a customer who asks follow-up questions, a colleague who needs a clear explanation, or a tutor who waits for your answer.

Choose something small enough to practise.

“Improve my communication” is a reasonable ambition, but a difficult instruction. “Help me explain a two-week delay to a customer without becoming defensive” is something you can rehearse this afternoon.

A useful task includes a situation, a constraint, and an outcome. For example: a client has asked for a discount; you can change the scope but cannot reduce your standard rate; you want to agree a next step without promising something you cannot deliver.

The same approach works for language practice. Choose a conversation you actually have: opening a team meeting in English, asking a supplier to clarify a delivery date, or explaining a project decision. A short, familiar situation lets you concentrate on expressing yourself.

Use fictional names and figures when setting up the exercise. The AI needs the shape of the problem, not your customer’s private information.

Make an attempt before asking for the model answer.

Write or say what you would do first. It may be clumsy. That is useful material: a vague opening, an unnecessary apology, a claim you cannot support, or a question you forgot to ask.

If the AI supplies a polished answer immediately, you lose that starting point. It becomes harder to tell whether you understand the approach or simply recognise a good version when you see it.

A field experiment with high-school mathematics students illustrates why this distinction matters. Students using a general AI helper did better on assisted practice but worse when later assessed without it. A version designed to give learning support largely avoided that negative effect.¹ This was a study of maths learning, not workplace coaching. It offers a reason to examine how a tool helps, rather than assume every assisted answer builds skill.

For your own practice, ask for questions or hints before a complete solution. A simple instruction is enough to begin: “Let me attempt this first. If I get stuck, give me one hint and wait.”

Give the session a beginning and an end.

Here is a 15-minute routine you can try with an AI conversation tool. The timings are a suggested structure, not a promise about how quickly a skill develops. You can shorten the exercise or spend longer on the difficult section.

Suppose the task is that discount conversation. Ask the AI to play a buyer with a limited budget who is also considering a competitor. Give it permission to resist your first proposal. It should stay in character until you finish, so you have to handle the conversation rather than receive coaching after every sentence.

Those figures are fictional. Replace them with a made-up situation that resembles the kind of decision you face. If the tool agrees too easily, add a credible concern, such as a procurement deadline or uncertainty about the value of the service.

Ask for feedback you can act on.

“You communicated well” does not tell you much. Neither does a score with no explanation. Ask the AI to point to the sentence, question, or decision behind its assessment.

For a sales exercise, you might examine whether you asked what was driving the price request before defending the price. For a meeting update, check whether you stated the decision you needed. For English practice, ask which phrase made your meaning unclear and why.

Keep the first review narrow. One change is easier to practise deliberately than a page of suggestions. Read the feedback against your actual words: AI can misread your intent, invent a criticism, or offer a confident explanation that does not hold up.

Then repeat the difficult exchange. Try a different opening or ask the missing question. You are looking for a change you can recognise in the conversation, not simply a higher score from the tool.

Take one change into the working day.

A practice session becomes more useful when you know where you will use it. Before you close the tool, complete this sentence: “In my next meeting, call, or email, I will…” Make it specific enough that you can tell whether you did it.

Afterwards, write a short note about what happened. Perhaps you asked a clearer question but still rushed to fill the silence. Perhaps the real customer raised an issue the simulation never mentioned. Both give you material for the next attempt.

Occasionally try the task without AI assistance. Explain the idea from memory, write the opening yourself, or rehearse with a colleague. That gives you a more useful check on independence than the number of practice sessions completed.

Keep a teacher in the picture.

AI makes it convenient to repeat an exercise. An experienced teacher or colleague can help you decide whether you are practising the right thing.

You might spend a week improving your pitch when the real problem is that you have not understood the customer. You might memorise more business vocabulary when your update would improve most by being shorter. A human who understands the context can challenge the task itself.

This is how we think about live learning, LONED Sales Coach, and LONED Language Coach at LONED. Take an idea from a class, practise a relevant situation, and return with a more precise question. Our guide to live and recorded courses looks at choosing the right format for each part.

The next step can be small. Pick a conversation you have been putting off. Practise its first two minutes. Listen for the part you want to change, and have another go.