Stuck at Work? How to Get Help From AI

You don't know what to ask

You open an AI assistant because you need help with a piece of work, then stare at the empty message box. You know something needs to change, but you cannot yet explain the whole situation or say what a useful result would look like. It is frustrating to get stuck before you have begun. So you type a quick question. The answer is generic, or the follow-up questions take you in circles. The work is still waiting.

For years, we have typed questions into Google and sifted through the results. If the question misses what we really need, so can the results. An AI assistant can respond to a one-line request in much the same way. It may even ask questions without getting closer to the goal, because you have not yet worked out what that goal is.

I have been there. I once asked an assistant for “a meeting structure, where each team reports their status regarding the progress of the quarterly goals.” That was what I could put into words. It left out the people involved, the decisions we needed, how often to meet, and what should happen when a goal was at risk. I had not worked out how to get people behind a change to the meetings either. I had seen colleagues struggle to get useful help too. Some said they would not use AI because it created more frustration than it solved.

I developed the prompt below so I could begin with the part I knew and work out the rest in conversation. Instead of trying to write a complete request, I let the assistant ask me what it needed to know. Each answer helped me understand the problem and the result I wanted more clearly.

Let the assistant ask

You can use the same approach for a customer meeting, figures in a spreadsheet, a process change, or a difficult email. Start with your rough goal. The assistant will ask one question at a time before it produces the work.

Open a new conversation with an AI assistant your organization permits you to use. Copy this prompt and send it as your first message:

You are my personal Problem-Solver.

I know what I want to achieve, but I may not know what information you need. Interview me before you start the work.

Ask one question at a time. First, understand what I want to achieve and what I want you to produce. Then ask only questions whose answers would materially improve the result. Help me think through relevant audiences, existing materials, constraints, risks, and what success looks like.

Adapt to my answers. If I am unsure, offer 2–3 options, explain the trade-offs, and recommend one. Point out important considerations I may have missed. Do not invent facts; label any assumptions.

When you have enough information, summarize your understanding and ask me to confirm or correct it. After I confirm, create a complete, usable output in the format that fits my goal. Include practical steps, owners, or follow-ups where relevant. Briefly explain how to use the result.

Start by asking what I want to achieve.

Work it out as you answer

The first answer can be as short as “I need to prepare a customer meeting.” Tell the assistant what you hope to have at the end, even if you are unsure. Then take the questions one at a time. You can say that a question does not apply, ask for options when you do not know, or correct the assistant when it has misunderstood you.

Before the assistant begins the work, read its summary of your goal. Correct anything that is missing or wrong. That check matters: a polished answer to the wrong problem will still leave you stuck.

Bonus: Dictate instead of typing

If you can, dictate your answers instead of typing them. Speaking helped me think out loud. As I heard myself explain the meeting problem, I noticed details I had not included in my original request. It also felt quicker than composing each answer at the keyboard.

The questions changed the task

The interview showed how much my one-line request had missed. The assistant asked what each group needed from the meetings. Each business representative wanted an update on their own topic, especially deadlines and decisions that needed their input. The managing director needed a view across teams. Team leads needed a lean way to report progress and see what might come next quarter.

As we talked, the purpose changed from collecting team updates to reviewing quarterly goals, spotting risks early, making decisions, and removing obstacles. Questions about our existing reporting exposed another problem: team leads were repeating information for different audiences. The principle we arrived at was to update information once and use it for everyone who needs it.

While answering aloud, I thought of a shared agenda that participants could fill in themselves. That took preparation off my shoulders and gave everyone a part in making the meeting work. The assistant then created a Quarterly Goals Review: Meeting Playbook with a proposed 60-minute meeting every two weeks, short written updates, topic-specific attendance for business representatives, and owners and follow-ups for open issues.

I use the playbook now. I have not measured the difference, but participants are happy with the cadence and the information they receive. We are more aligned, find risks and dependencies earlier, and spend less effort in meetings. When an issue cannot be resolved in time, it stays on the list with a clear next step.

Try it on your next task

Think of one piece of work you have been putting off because the request is hard to formulate. Paste the prompt and tell the assistant what you know so far. Answer its first question; you can work out the rest together. When it produces the result, check the facts and make sure it fits the people and decisions involved before you put it to work.