AI Can Build Your Pitch Deck. It Can’t Tell You What to Say in It.

The core problem: AI likes sensible slides

This is useful. And it is also where things start going wrong.

Because pitches rarely fail because slide 17 doesn't make sense.

They fail because slide 17 doesn't need to exist. And probably slide 18 either.

AI is very good at finding something sensible to say about the information you give it. It is much less comfortable deciding that something perfectly sensible should be left out.

But that's a huge part of writing a good presentation.

I don't want every fact we know. I want the facts I need to make the argument.

And I don't want every slide to have equal importance.

A pitch should change how the room feels

There are lots of perfectly logical frameworks for building presentations.

Situation. Problem. Insight. Solution. Proof.

Nothing wrong with them.

But I tend to think about pitches differently. I want something closer to an emotional rollercoaster.

Start somewhere relatively comfortable. The client's brief. Their business. Something everyone in the room broadly agrees with.

Then go down.

Make the problem worse. Show the tension. Demonstrate why the obvious answer isn't enough. Ideally, at some point the audience should feel slightly more uncomfortable about the problem than they did before the presentation started.

Then give them hope: there is a way through this.

And then reveal the solution.

That's normally the emotional high point. The moment where the things you have shown before suddenly come together and the answer feels both new and, with hindsight, strangely obvious.

After that you can come down again. Explain it. Prove it. Show how it works.

You can't stay at the emotional high for 40 slides, anyway.

People don't want loans

We once pitched for a bank where the brief centred on a feature: customers could apply for a loan by video call 24/7.

There is a very obvious presentation you could build from that:

People expect convenience. Digital behaviour is changing. Consumers want services on demand. The bank is available whenever you need it.

Perfectly plausible. And boring as f*ck.

And, if we are honest, not really true, either.

Because deep down we all know a very obvious thing:

People don't want loans. They want the things loans allow them to buy.

A car. A kitchen. Whatever it happens to be.

And that changes the problem.

Our strategic idea was to put the bank next to the things that might make somebody need a loan. Even in television. If a car ad appears, follow it with the bank.

Suddenly the 24/7 video-call feature has a role.

The argument isn't our loan application is convenient.

It becomes be there at the moment something makes people need the money.

Now imagine the presentation.

Everything before that thought should make the problem feel increasingly clear. Then comes the reveal. Everything afterwards explains why it works and what you do with it.

This is why I end up rewriting generated decks

When I use AI presentation tools, I frequently end up rewriting most of the presentation.

Not because the individual slides are necessarily bad. Many of them aren't.

The problem is that the deck has been assembled rather than argued.

There is too little tension. Too little understanding of which information matters more than other information. And too little courage to simply leave things out.

Again, I don't think this makes AI presentation tools bad.

It means we're asking them to solve a different problem.

By all means use AI to structure material, explore ways to visualise something, produce first drafts or kill some of the hours spent nudging boxes around PowerPoint.

I'll happily take that.

But don't confuse producing the deck with deciding what the deck needs to say.