AI Applications · Knowledge Work

Tidy the Mess Before Consulting: Let AI Turn Chaos into a Question

Have you ever wanted to talk something through with someone, but couldn't even explain what was blocking you? Before you open your mouth, collect your materials in one place and let AI turn the mess into a question you can actually bring to the table.

A white-cat consultant, an AI assistant inside a laptop, and Mika organizing a messy pile of notes on the desk
Collect your materials in one place. Let AI turn the mess into a clear question first, then bring it to the conversation.

Often, what we lack is not a consultant. The real obstacle is that we cannot articulate the problem clearly enough to hand it to someone else. This article covers a method anyone can use: collect your materials, the other party's materials, and every past discussion in one place, such as a ChatGPT project folder, so AI can turn the chaos into a question someone else can actually receive, before you go to a consultant, your team, or a meeting.

👥 Who this is for
  • People who wanted to talk with a consultant or coach and were told to "organize their thoughts first"
  • People who meet regularly with the same team or work on the same project and have to start from scratch every time
  • People who want to use AI to clarify their thinking but don't know where to begin
🎯 What you will get
  • A "problem summary" you can bring directly into a conversation
  • A method for unpacking a problem into three layers: the surface question, the solution you have already assumed, and the deeper question you haven't surfaced yet
  • A prompt you can paste directly into ChatGPT

The consultant told me to "organize my thoughts first," but organizing them was exactly what I couldn't do

I once went to a consultant to talk through a business problem. I poured out too many concerns at once, scattered in every direction. The consultant couldn't catch it all and asked me to come back once I had organized things.

That sounds reasonable, but my real difficulty was that I couldn't organize things in the first place. If I could have done that on my own, I probably wouldn't have needed the consultant as badly. It became a loop: the more I needed help, the less I could produce something the other person could work with.

You may have run into the same thing. You want to discuss a work situation or align with your team on a project, and the moment you start talking, five projects, three constraints, and two threads of old context come spilling out. The other person is lost, and you get more tangled the longer you talk.

Turning a mess into a question: a real before-and-after

Here is an example from my own experience. I was stuck on "students aren't understanding my AI course," and my brain jumped straight to "maybe the curriculum isn't complete enough." When I framed the problem that way and brought it to someone, all I got back were suggestions to fill in more curriculum. Later I tried something different: I let AI spread the mess out and unpack it layer by layer, and the problem looked completely different.

Raw, messy version The question I instinctively reached for

"I feel like students aren't understanding my AI course. Should I make the curriculum more complete?"

After AI unpacking Three layers revealed

The surface issue is the curriculum. The solution I had already assumed was adding content. The deeper problem may be that students don't know how to fill in missing background knowledge, and don't know how to judge whether an AI's answer is reliable.

Consultant can engage The direction becomes clear

Hold off on expanding the curriculum. Design a small practice that lets students try filling in background knowledge and verifying AI outputs instead.

Here is the difference AI turned one tangled sentence into a question with distinct layers. A human consultant, receiving that structured version, could actually engage and get to the point. When you hand someone a mess, they can only get messy with you. When you hand someone a question with its layers already separated, they can actually help.

Why we so often can't produce a clear question

One common reason is that before we even open our mouths, we have already translated the difficulty into a solution-shaped question in our own familiar terms. "How should I teach this?" "What's missing from the curriculum?" "How do I get everyone up to speed?" When a question arrives in that form, AI will dutifully follow your framing and reinforce it, rather than helping you step outside of it.

Something easy to overlook What we most readily miss is the automatic, unthought judgment we make every day. Things that are obvious to you may not exist at all for someone else. That's why spreading the problem out and asking AI "have I locked in an answer too soon?" is often more useful than asking directly for a solution.

Anyone can do this: open one project folder and collect your materials

The simplest version requires no special setup. Open a ChatGPT project folder and keep adding relevant materials to it: your own documents, the other party's documents, and if you meet regularly with a particular consultant, team, or recurring project, add the notes from every meeting and every discussion into that same project.

That gives AI the context it needs. The next time you get stuck, you don't have to recount an entire month of history from scratch. It already knows the background and can help you turn a mess into a question before you even start explaining.

The same approach fits many situations: recurring consulting sessions, regular team meetings, ongoing project collaborations. When materials accumulate in one place, AI can keep helping you organize them as you go. When you're stuck, paste in this prompt:

I have a messy problem. Please don't rush to give me solutions. First, help me clarify: 1. What am I actually trying to solve? 2. What answer have I already assumed? 3. What background am I leaving out? 4. If I were going to bring this to a consultant, how would I say it in three sentences? Please give me: a three-sentence summary, three questions you want to ask me, and three questions worth bringing to a real person.

Going further: let an AI agent read your entire context

If you want to take this further, there is an agent version. Instead of just uploading materials to the ChatGPT web interface, you can let an AI agent that has access to your entire knowledge base and work journal do this work. I've actually tried handing my laptop to a consultant and asking her to talk directly with Codex on my machine.

The difference is context. My knowledge base, work journal, and the full history of recent projects are all on that machine, so every follow-up question the consultant asked, the agent could immediately pull in the relevant context and add it to the conversation. I didn't have to reconstruct anything from memory. The resulting question was far more complete than anything a single conversation could produce, and the follow-up questions went much deeper.

The concept is what matters The agent version is just a more complete and automated way of doing the same thing. You don't need to set that up first. Start by building the habit of collecting materials in one place and letting AI turn them into a question. The ChatGPT web interface is enough to get started.

AI organizes. People judge.

One thing worth being clear about: AI's role here is to organize and translate. The final judgment still belongs to a person. These two things each have their own place.

AI is suited for Organizing and translating

Spreading out the chaos, listing possible blind spots, unpacking the problem into layers, producing a version of the question you can bring into a conversation, simulating different angles to rehearse your thinking in advance.

Human consultants and teams are suited for Judging and bearing responsibility

Deciding which direction is more right, pointing out assumptions you haven't seen, unpacking real-world situations, and being present with you as you make a choice. These are things only a person in the room can hold.

In short: let AI help you make the question complete, then leave the judgment to a person.

Before your next conversation, try it once

Open a ChatGPT project folder, add the materials and past discussions related to the topic, paste in the prompt above, and bring the "problem summary" AI produces into the meeting. You'll find that the conversation starts from a much clearer place.

You don't need a sophisticated tool. Building the habit of collecting materials in one place and letting AI turn them into a question first is enough to get started.

Want to practice this method together?

I share more ideas for connecting AI to everyday work on my LINE community and Threads. Come say hello.

Join LINE Community ↗