How to Use AI to Get Up to Speed on a Topic You Know Nothing About — Fast

Someone sends you a document about transfer pricing and wants your thoughts by Thursday. Or your company is about to be acquired by a private equity firm and you have no idea how that process actually works. Or you've been pulled into a meeting about supply chain logistics for an industry you've never worked in, and you'd like to not sit there nodding blankly for an hour.

These situations come up constantly in working life — the sudden need to understand something you have zero background in, on a timeline that doesn't allow for a proper course or a book. This is one of the places where AI tools are genuinely, practically useful in ways that aren't overstated. Used well, ChatGPT or Claude can compress what would otherwise be two days of reading and Googling into a focused two-hour session that leaves you with real working knowledge.

But the approach matters quite a bit. There's a significant difference between asking AI a topic question and using it as a structured learning tool. Here's what that difference looks like in practice.

Start With the Shape of the Topic, Not the Details

The most common mistake when using AI to learn something new is going straight to the specific question you need answered. That feels efficient, but it usually isn't — because you don't yet have the frame to understand the answer properly.

Before you ask about the specifics, ask AI to give you the shape of the topic. What are the main concepts? What are the moving parts? What does someone who works in this area spend most of their time thinking about? This gives you a mental map before you try to navigate the territory.

In practice, that might look like this. You type into Claude or ChatGPT:

That "don't assume any prior knowledge" instruction is important. Without it, AI tools often pitch their explanations at a level that's just above where you are — using terms that seem like they should be obvious but aren't. Telling it your actual starting point gets you a genuinely accessible explanation rather than one that's accessible to someone who already knows 60% of what you need to know.

Build Up in Layers, Not in One Pass

Once you have the overview, resist the urge to immediately ask about your specific situation. Instead, spend a few exchanges building up the topic in layers. Ask for one concept to be explained more deeply. Ask what you're still confused about after the first explanation. Ask what you'd need to understand to have a genuinely informed opinion on the topic.

This layered approach works better than a single long prompt for a simple reason: you learn what you don't know as you go. After reading the first overview, you'll have real questions you didn't know to ask before. Those questions are usually more productive than whatever you would have typed at the start.

A useful follow-up pattern after any AI explanation is:

That second question is particularly useful. AI tools are good at anticipating the gaps in someone's understanding and surfacing questions a novice doesn't yet know they have. It often points you toward exactly the thing that would have tripped you up in the actual meeting or conversation.

Ask AI to Teach It to You a Different Way

Concepts that don't land on first pass often land immediately when explained through a different frame. If you read an explanation and it doesn't quite click, don't just re-read it. Ask for a different angle.

Some approaches that work well:

Ask for an analogy. "Can you explain this using an analogy that has nothing to do with finance?" often produces the version that finally makes something stick. Good analogies don't just simplify — they give you an intuition for how the underlying logic works, which means you can reason from first principles rather than just remembering a definition.

Ask for a concrete example. Abstract explanations are harder to retain than specific ones. "Can you walk me through a real example of how this actually plays out?" turns a concept into a story, which is far easier to remember and explain to someone else.

Ask for the simplest possible version. "If you had to explain the core of this in three sentences to someone who'd never heard of it, what would you say?" This strips away everything non-essential and leaves you with the kernel you actually need.

In my own experience, it's rarely the first explanation from an AI tool that makes something land — it's the third or fourth, after asking for it a different way. Most people stop too early.

The Verification Step You Can't Skip

AI tools — ChatGPT, Claude, Gemini, all of them — occasionally state things that sound authoritative and are simply wrong. This is known as "hallucination": the AI generates a confident-sounding answer that isn't accurate. It happens less often on well-established topics, but it does happen, and when you're learning a new subject you won't always recognize a plausible-sounding error.

The practical rule is this: use AI to build your mental map and understand the concepts, but verify specific facts, figures, regulations, and claims through primary sources before you rely on them in anything important. If you're learning about transfer pricing rules and AI tells you a specific OECD threshold or a particular court case, look it up. The concept the AI explains will probably be right. The specific detail may not be.

For this kind of verification, Perplexity is often more useful than ChatGPT or Claude because it shows you its sources inline — you can see where the information came from and go check it directly. A reasonable workflow is to use Claude or ChatGPT for the conceptual learning and Perplexity to verify specific facts.

Preparing to Sound Like You Know What You're Talking About

Once you have a solid grasp of the topic, there's a specific preparation step that is genuinely valuable before any meeting or conversation where you'll need to use this knowledge.

Ask AI to run a practice conversation with you. Something like:

This is a different kind of learning than reading explanations — it's active retrieval, which is significantly more effective at building durable understanding. Trying to engage with the concept in real time, rather than just reading about it, reveals what you've actually absorbed versus what you only think you understood.

You can also ask AI to generate the three or four questions a thoughtful person in this meeting would be expected to ask, and to explain the reasoning behind each one. Walking in with those questions — understanding not just what to ask but why it matters — is the difference between looking like you're following along and looking like you belong in the room.

What AI Can't Give You Here

There is a ceiling on what this kind of AI-assisted crash course can do, and being realistic about it prevents overconfidence.

AI can give you a working model of a topic — enough to follow a conversation, ask reasonable questions, and understand the answers. It cannot give you the judgment that comes from years of working in an area. It can explain what transfer pricing is and how it works; it cannot tell you which arguments tend to actually hold up in disputes, or which approaches specific regulators tend to scrutinize, or what the experienced practitioners in your industry think about the current state of the rules. That knowledge comes from people, not from AI.

The appropriate goal for a two-hour AI-assisted learning session is informed competence, not expertise. You want to understand enough to engage meaningfully, ask good questions, and recognize when something being said sounds wrong or unusual. That is a genuinely useful thing to walk into a meeting with. It is not the same as actually knowing the subject deeply, and it shouldn't be mistaken for that.

Used with that realistic expectation, this is one of the more consistently valuable things AI tools do. The knowledge gap between "I've never heard of this" and "I can hold my own in a room where people talk about this" used to require weeks of reading. It now often requires a focused afternoon. That is a real change in what's possible — not for the expertise, but for the on-ramp to it.

The next time something lands on your desk that you know nothing about, that's not a problem to manage around. It's two hours of work, and a tool that's free or close to it, and then you're in the conversation.