Introduction

I want us to do a research project, which should last about 20 to 25 minutes.

Defining the Research Goal

And I want us to imagine that we need to research a new market. And we need to present this market to our senior leadership in a corporation to decide whether we would want to entry or invest in the market or not.

And then I will use a bunch of tools to do some research, come up with a presentation, perhaps a video. Let's see how far we get that we can share with our senior leadership to help them guide the decision.

Tools And Audience Poll

I will use, let me see a little bit how much you know about some of the tools that I will be using. Who has used, let's start from scratch.

CharGBD, who has used CharGBD? Who has used Gamma? Half.

Who has used Suno to create music? And who has used notebook.lm?

OK. A mix.

Hopefully, you will all learn something today. Seems so, maybe apart from Vincent and a few others.

And before we start, who has used Super Whisper? Who has heard of Super Whisper?

I've used the competitor. OK. Well, you're familiar with the concept.

Selecting the Market

Now, I said we want to research a market, let's come up with a market. What market we want to look at? Could be anything.

Bouldering gyms.

That might be, we're stretching it, it might be too specific. It could be, like, gyms could be, perhaps, like gym chains. Thinking about there needs to be some publicly available information.

That's the limit. Because otherwise, the research will be, we will struggle with the research.

So gyms could be one. Anything else?

Real estate? That's probably too broad. So it needs to be. Yeah, it needs to be a segment of real estate.

So gyms could be a segment of real estates. For what specifically? Okay, let's go with that.

I was planning to use, have you heard of Napkin? Yeah, no, I was planning to do it on the whiteboard. Let's, I mean, because napkin wasn't working, it was down, but let's use napkin so you see another tool.

Finalizing the Target: Fleet Management Solutions

So fleet management solutions.

Framing Executive Questions

What do we want to look at in the fleet management solutions market? What are some questions that we want the executives to answer?

This solution is present in Palestinian market. And Palestinian company, because of the war, they want to come here, maybe in Europe, and develop this product.

So they wanted to know other solutions, functionality of the solution, price and performance of the solution. And the differentiation model, yeah, for entry in the market, what they can offer.

OK.

Key Metrics: Pricing, Competitors, Growth

Price.

Main competitors. What else?

Market growth. Market growth, OK.

Anything else?

Strategic Considerations: Entry Viability and Barriers

In general, for them, is it worse entering the market? How hard it is to enter the market? Barriers to entry.

Visual Brainstorming with Napkin

Now, what Napkin does, and we see we have provided very little context. Just select the text. Then it would generate some infographics for you.

Of course, not working. Typical.

You have troubles as well? Yeah, I think it is.

It was down earlier, which is why I was planning to use the... Okay, we managed.

Using Visuals to Structure Thinking

And so if you're not a visual person, Like, I'm not a visual person at all. I'm a man of words and concepts.

This is really useful even just to break the ice and think about how you could present these ideas visually. I really like the spiral, doesn't it, for this one?

Let me copy it down so you can see the full thing. Let's pick one just for the sake of argument.

From Whiteboard to Prompt

Now, imagine we had actually put this on a whiteboard like I was planning to do.

Crafting a Research Prompt with GPT

Now we're going to go to GPT, and we will ask to take all of this information and come up with a research prompt that we can run.

Imagine you are an experienced consultant from a prestigious consulting company.

I'm here with my team and we have come up on a list of factors we want to do some research on to inform the entry in the fleet management solutions market in Europe. On the back of this research that we have done, can you come up with a deep research prompt that we will use to conduct research on an LLM tool?

Okay, you see I'm not a native speaker, so I need to double check this. Mine works fairly well.

All I do is I put the image that I just put together. It's uploading. Then we wait, hopefully not too long. Okay.

Obviously, we could have written this ourselves, but it would have taken a lot longer, and it also wouldn't have been optimized to the way in which an LLM, so a large language model, would understand.

So what do I do with this?

Running Parallel Deep Research

I take this, and I run it in parallel in two research tools. One of them is Perplexity. Have you heard of Perplexity before?

Yes, a few people have, a few people have not. And I'm using the labs function and for those who have used perplexity before, what the labs function does, it creates separate assets that you can work on in parallel.

Setting Up Perplexity Labs and Gemini

The only thing I will change here, focus I will make it a bit shorter because we don't have that much time focus on five pages and then we run this now we do the same thing in Gemini always use the deep research function by Gemini okay few people and

Why Parallelize: Cross-Checking and Speed

1So the reason I run them in parallel, then we could cross-check them, put them together. Gemini is typically more powerful, but it would also take a bit longer to run.

Now, what do I do while I wait for perplexity?

Yes? Like this? OK. Thank you for raising that.

What Perplexity Labs Produces

Perplexity? So perplexity is sort of in between, between Google and ChartGPT.

So it would run a series of parallel searches. And then it would save the results and summarize them. in a document or in any output that you specify.

The advantage of using this labs function that I have now is that you will see you have different assets. So you don't just have the PDF document, which is what Gemini would do or ChartGPT would do, but you also have a chart if it creates a chart. You also have a specific summary and infographic that it has elaborated for you.

Does that make sense? Any more questions on this? Those that have used perplexity, have you tried the labs function before?

Curious. OK.

Interlude: Creative Demo with AI Songwriting

Now, what do we do as we wait for this? I mean, we could go for a coffee, for a walk.

But what we're going to do is I want to show you the song generation. functionality.

Generating Lyrics Aligned to the Theme

So what we will do back on ChudgyPT, imagine you are an experienced songwriter. We are in a church right now. So I would like you to give me the text for a folk music song that would be nicely played in a church. Let's see what it comes up with.

Okay. Not bad. Yeah.

Yes, but I want it to be more catered Yes, but I wanted to be linked to fleet management solution. Yeah.

Now we're talking. Not bad.

Can you read? Not bad. All right.

What do we do with this? And you can also use it.

From Poems to Songs: Structuring for Music

So I write poetry in my spare time sometimes. I basically ask to create a text out of a poem that I've written. Because of course, the poem is not maximized to be used for music. But what ChartGPT would do would basically give it the structure of a song, something that I wouldn't know how to do intuitively.

I basically ask, maybe if we have time, I can do it later.

Yeah, I know, I know. I was trying, but I didn't quite manage to because it was too zoomed in. Thank you for spotting that.

Where's your before? Remind me.

OK, just a few people. There we go.

We've got our song.

actually now we want to have we want to have the lyrics we lost them of course so we go back I'm struggling with the copy paste. Why am I struggling here? Not sure.

Ah, we managed. Just took a little bit of time. Wasn't patient enough.

So we have the, we paste it multiple times, great. This is what happens in a live demo. Sometimes you get things wrong.

I might just refresh the page, right? Okay, beautiful. We have our text.

Producing and Reviewing the Track

Now we set country music. Let's see what happens and if we like it.

We gather on the journey Through night and through the rain The wheels are turned together We'll carry us again It is about fleet management solution. It's country music, and it's quite churchy.

Sorry? Yes. I should get a cap. And now with the song as well, right?

Okay, where are we? And obviously, I mean, this is just, we were sort of playing with it now, but you can use it for poems, you can use it if you want to create a jingle in a client meeting to open up the discussion. There are many different ways in which this could be useful.

You had a question? No, no, no, okay, sorry.

Now.

Hoo hoo.

Reviewing Research Outputs

Perplexity has gotten back to us. It's quite a bit.

And now you see the labs function. You've got these separate assets that you can use. You can pick them individually, which is something that I'm not sure you can do in the deep research function.

Exporting and Preparing Summaries

Now, what do we do with this? I will export it as a PDF.

Turning Research into a Presentation

We go back on ChatGPT. We are back with the research.

Prompting for a Slide Outline

Imagine you are an experienced presentation maker.

Now I want you to summarize

There is also the research in an outline of about 10 slides for a presentation that I can present to my senior leadership. After each slide, please put three dashes because I need to export this to gamma.

Okay. And we want, let me rephrase this.

I attach the results of the research. Let's see what it comes up with.

Model Choices: Instant vs Thinking

By the way, here I'm using the instant model.

So if I were on my own, I would be using the thinking model, which is a bit more powerful and more detailed. It just takes a bit longer, and I don't want us to have to wait.

Who is experienced with different models? Like, let's try the thinking model. It used to be called O3.

Yes. A few people. OK.

Yes, which is this. You will see why.

This is because I'm being lazy. Because now, obviously, I would normally spend some time revising this.

Editing in Canvas

Yes? Why don't you use Canvas? I was about to say, I would normally spend some time revising this in the Canvas.

Can you open this in a canvas please for me? And so

This is quite handy, the canvas, because then I would have the chat on the left and the text on the right, and I would be able to continue chatting with ChatGPT and ask to make changes to the text. It's like a side-by-side window, which is quite common in programming, but not really, not yet common on the non-technical side of things, I guess.

Now, let's assume that This works, and obviously I would normally spend a good amount of time double checking this and ensuring that it's all aligned, but I want us to have enough time to play with Gamma.

Building Slides in Gamma

So Gamma allows you to create presentations from scratch through some agentic-like interface. You have different options. We're going to paste the text in.

Importing the Outline and Auto-Segmentation

We want the presentation. from an outline.

Now you will see why I have the lines there. All I did was to copy and paste that. Because when I've got the lines in, then it splits the text into different cards. So it already knows how we want the slides to be segmented, to be divided.

This is gamma. This is gamma. Now you can, and I will show you how to get there.

That's what we did. This is how gamma looks like. And I went on just creating a new presentation.

Styling, Audience, and Templates

Now, you can choose the amount of text that you want. You can choose your audience, the tone. And you can choose your image style.

You can import your own templates. We have our own Mindstone templates.

Let's go with, let's keep it simple for now. Now, let's see what it comes up with.

Bless you.

Agentic Edits and Auto-Generated Assets

So just on the back of that, of this outline, it would generate the full slide deck, and we would then be able to just make changes pretty much as I was doing with ChartGBT, just by charting with Gamma. It would come up with charts. Slides. Timelines.

Barriers to entry. And I mean images that would match the time, that would match the theme that we have chosen.

Can you see the screen well? Is it showing in the back?

Now I want to show you a few things, like how you can make changes quickly. I don't think this is a timeline. This looks more like a list of bullets. Can you please change this?

And that's where the agentic function comes in. And you see that I took my prompt. and it replaced that timeline with a list, pretty much as I asked.

Not sure I like it, but it's better than what we had. Don't like the fact that there is a gap there.

Wrap-Up and Time Savings

All right, so you see how in, let's say probably 25 minutes, we moved from just a list of things to a presentation. Probably did.

If we think about the amount of time that would have taken us to do this, we're probably talking about two weeks. Of course, in a normal workplace setting, I would spend time revising the outputs at each stage. But this was just meant to give you an idea of how you can combine these tools and how you can use voice to speak to these models.