Use ChatGPT to Attack Your Presentation Before Your Boss Does

What makes you nervous before a presentation is usually not the deck.

It is one question you know is coming, and you do not have an answer for it yet.

I lived with that feeling for years. Meeting rooms at Tokopedia, Hijra, Flip, and now leading four product teams at a MENA superapp. Deck clean, numbers complete, talk track rehearsed three times. Then one director asks where that assumption came from, and the whole thing falls apart.

The problem is simple. You practiced delivering. You never practiced being attacked.

Why normal rehearsal does not help

The most common form of practice is reading your slides until you sound smooth. That trains your mouth, not your material.

But presentations almost never collapse during the presentation. They collapse in Q&A. That is the part you cannot memorize. That is the part where someone senior pokes at the one hole you quietly knew was there and hoped nobody would notice.

There used to be exactly one fix: find someone senior enough and blunt enough to tear your material apart in advance. That is hard. Those people are busy, and they usually give you their comments on the day itself, in front of everyone.

Now you can summon a synthetic version of that person whenever you want. This is the most underrated use of ChatGPT for presentation practice. Not writing your slides. Attacking them.

Step 1: Give it the room, not just the deck

The usual mistake is pasting slide content and asking what it thinks. You will get something polite and useless.

The model needs to know who is in the room, what they care about, and what decision you are asking for.

I am presenting to a Head of Operations and a Finance Director. The goal is budget approval for one new tool. The Head of Ops cares about team workload. The Finance Director cares about payback period. I am going to paste my outline next. Do not comment yet. Wait until I say I am done.

Then paste the material.

Step 2: Make it a hostile executive, with a motive

Telling it to be a tough boss is too vague. You get generic questions back.

What makes it sharp is a role plus a history:

Now act as a skeptical Finance Director. You have been burned twice by projects that promised savings and delivered nothing. Give me the 10 most uncomfortable questions you would throw at me in that room. Rank them by how badly they would expose me as unprepared. Do not answer them.

That last instruction matters. If you let it answer its own questions, you are just watching. You want to sweat now, not on the day.

Step 3: Go back and forth, not one round

This is the step almost everyone skips. Answer the questions one by one, type your real answer, then:

Score the answer I just gave, from the perspective of that Finance Director. Which parts are still vague? Then hit me with a harder follow-up.

Three or four rounds in, you will hit a question you genuinely cannot answer. That is not a failed rehearsal. That is the whole point of the exercise.

Step 4: Dig out the silent assumptions

Last, tell it to stop asking and start dissecting:

Stop playing the executive. Be an auditor now. List every assumption I am relying on but never state explicitly in this material. Flag the ones that would break first if someone asked me for the source.

I almost always find two or three assumptions I thought were obvious. They were only obvious inside my own head.

Two things you must not do

The model does not know your office politics, your project history, or your internal numbers unless you hand them over.

So, two hard rules.

First, do not paste sensitive company data into a tool your employer has not cleared. Check the policy. If you are unsure, convert the figures into ratios or use dummy numbers with realistic proportions.

Second, never ask AI to invent supporting numbers for you. If you do not have the data, the honest answer is that you do not have it yet and you are taking it as an action item. That sentence is far safer than a fabricated figure that gets caught seven minutes later.

Why this matters for your career

Saving companies money is an old habit of mine, long before AI became fashionable. More than $4 million per year, across a mix of initiatives: process fixes, cost optimization, product decisions that were simply more efficient. Automation was only one of them.

None of those numbers happened because I had a good idea. They happened because I convinced a room full of skeptics. An idea that does not survive the meeting is worth zero.

Back then, that kind of leverage required position, an engineering team, and expensive systems. Today AI is the sharpest tool for the same habit, and it can be taught to everyone on your team.

I see it from the other side now. I am not an engineer, but the 8-channel content distribution system behind my personal brand is something I built and operate alone, with AI. Same principle as before. Different tools.

Rehearsing with AI is a small version of that. You move the moment of I am not ready out of the meeting room and onto your own desk, the night before, for free.

If you want to learn to think and work with AI this way, I go through it regularly with members of AI Circle.

If what you want to level up is a team or a company rather than yourself, the internal workshops are on the corporate page.