Copilot and ChatGPT: Optimizing AI Assistance for Your PCs

Key Takeaways

1AI chatbots like Copilot can assist in diagnosing computer issues, but their effectiveness varies.

2To obtain useful answers, it is crucial to ask clear and precise questions to the chatbots.

3Asking for alternatives and confidence levels can reduce AI overconfidence in their responses.

Why it matters : Better interaction with AIs can enhance IT troubleshooting and reduce user frustration.

The Challenges of AI Chatbots in IT Troubleshooting

Faced with often frustrating and error-prone IT troubleshooting sessions, one option is to seek help from Copilot, an AI assistant. One of the most obvious uses of an AI chatbot is to diagnose issues with your PC or Mac and propose solutions to fix them. However, the results obtained with these tools are, to say the least, uneven. Sometimes, the chatbot clearly identifies the problem and offers a logical solution. But there are also times when the responses are disappointing, or even incorrect, despite the AI's apparent confidence. The question is whether the issue stems from the user, the AI, or a miscommunication between the two.

Improving Communication with AI

One solution is to ask the chatbot itself how to phrase questions to maximize the chances of receiving relevant answers. Asked about this, Copilot, which is based on the ChatGPT model, offers very enlightening advice. Its recommendations also apply when you are trying to resolve an issue with a human technician.

Crafting an Effective Troubleshooting Request

Copilot introduces itself as "Microsoft Copilot, based on the GPT-5 chat model." This means it is designed for conversational reasoning, troubleshooting, and guided workflows, like diagnostic discussions. It says it is fully capable of helping to resolve Windows issues, especially if the user provides precise information. It can interpret error messages, identify likely causes, and suggest steps to take. It notes, however, that it performs better when the problem is described clearly, with relevant details.

The crucial element in a request, according to Copilot, is to start with an accurate description of what is happening and what was expected. Saying "My PC is slow" is vague. In contrast, "My Windows 11 PC freezes for 10 to 20 seconds when opening File Explorer" is actionable information.

The chatbot suggests a simple and effective format to follow:

Le brief IA que lisent les pros

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisissez votre rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

Problem: Describe what is happening.
Error messages: Provide the exact text or codes.
Recent changes: Mention updates, installations, or hardware modifications.
System details: Indicate the version and type of device.
What I've tried: List the steps already taken.

What can make a big difference, it adds, is mentioning recent changes, such as updates or driver installations, as many issues arise after such modifications.

Managing Chatbot Overconfidence

If the first response does not resolve the issue, Copilot advises treating troubleshooting as a conversation: add details, share results, and ask follow-up questions. Each step helps refine the cause.

The chatbot can also be very confident, even when the answer is incorrect. Copilot considers this a valid observation and explains that this behavior can be influenced by phrasing questions differently. "I am designed to be helpful and decisive, which can come across as overconfidence," it acknowledges. The user can explicitly ask for uncertainty and alternatives from the outset.

More specifically, Copilot recommends asking for likely and unlikely causes, along with the level of confidence for each. This immediately changes the tone, and the chatbot starts to nuance its responses. It is also possible to ask it to explain its reasoning before giving a recommendation, which makes it easier to spot weak assumptions.

It can even be forced to be less confident, by explicitly challenging the forthcoming answer. This involves questions like "What could be wrong with your response?" According to Copilot, this pushes it to switch from "solution mode" to "analysis mode."

Going Beyond 'Technical Support' Mode

Nothing prevents a user from discussing a problem with the chatbot as two competent peers. For complex issues, Copilot finds it helpful to frame the interaction iteratively: tell it not to jump to conclusions and to ask for more details if necessary. This allows it to pause and adapt to new information.

A few precautions are in order: avoid executing commands you do not understand, be cautious with registry modifications, and verify any steps that could affect data or system stability.

As for the best way to use AI for troubleshooting, Copilot sums it up this way: "Consider me a competent assistant, not a substitute for good judgment." The user provides the evidence, and the chatbot helps interpret it and suggests the next steps.