Me and AI

I am very interested in learning. Continuous learning. Although I am not the youngest, my stamina, willingness, and capacity to learn have not diminished; in fact, they have grown even more over the past few years. Part of it was/is the AI-driven revolution.

I started using Perplexity 2 years ago and learned the basics of AI and prompting. I have used it mostly for research and data/information consolidation. 6 months ago, I moved to Claude and recently to ChatGPT. I have started experimenting with many things. From Dashboards to Apps, from Documents to Slides, you name it. Step by Step. Recently, I have decided to move to my domain and see what it can do with my Data Storytelling passion. I tried to work on a Data Storytelling Slide.

Can it completely replace me, or do we work together side by side?

So I was thinking about where to start, and I came back to what I have already done in the past. My Course for beginners: The Curious Beginner.

I give it a simple background – based on the Background step – and see if it can help me to create something I am happy with. Not a 100% slam dunk, but pretty good enough. You can be the judge yourself.

I strongly recommend that you take a look at the whole training. No prerequisites are required, just a little bit of time and your willingness to change your career.

“The secret of getting ahead is getting started.”

The Beginning…Step 1

I have asked Claude to look at this table and the background of the story of Aisha and her problem statement around the drop in show-ups for her “Handling difficult customer” training. I asked it to create a data storytelling slide like Bain, McKinsey, or BCG consultants do. I asked it to look for the best framework to start with. And yes, this isn’t about having the best, most efficient prompt from the beginning. It’s about experimenting. But if you do ask, I tend to use the COSTAR framework for effective prompting.

Pure prompt looked like this:

I want you to create a data storytelling slide about this situation. It needs to look Executive and be done like a top consultant from Bain, McKinsey, BCG, or others. Search online for the best framework for a storytelling slide. Ask me if you need anything else to proceed.

It asked me 3 questions:

• Shall I correct the row 7 as it shows 150 sign-ups?
  • It proposed to changeed it to 15
• Who is this slide for?
  • I told it its for Ben and Mel for joint decision
• What should the story focus on?
  • Why the signup drops and include show up rate in your story too

It gave me the following output:

This is what came back first. It plotted almost everything on the slide. Both sign-ups and show-ups. A capacity line. Four colors. It wasn’t wrong; every number was technically correct, but it answered “what is in the table” instead of “what should Ben and Mei do.” That’s the usual trap for any analyst: showcase everything you’ve done because it “should” build your credibility. But it does not when presenting to leaders.

I didn’t like it much because it was too cluttered. It did not follow Edward Tufte’s Data-Ink Ratio.

My feedback was following:

Great attempt – but I think the slide is too busy – too many elements (just thinking about Tufte’s Data Ink Ratio). You can collapse the 2 charts into averages for those periods, Before and After, and clearly show the 42% decline already in the chart. Too many colors on the slide – let’s stick to 2 max and make the slide white (let’s not be afraid of white). Use the colors wisely, and if there is an opportunity to connect the colors already in the headline (e.g., “42%” to the chart showing the 42% decline, we can use the same color across the slide to guide the audience from the headline to the chart to the comments, etc. Let’s give it a try.

Again, I just followed basic, simple concepts of Data Storytelling and Data Visualizations and gave it another chance to succeed :-).

It came with this:

---

So the second version threw almost everything away. White background instead of cream. Two bars instead of twenty. One accent color instead of four. The 42% moved into the headline itself, in orange, so a reader’s eye had somewhere to land before it even reached the chart. This is where color stopped being decoration and started being a guide — the same orange on the number, the bar, and the arrow between them, so the eye walks headline to chart to explanation without being told to. Strategic color is the whole theory behind that move. But it put the orange in places where it shouldn’t be. I needed to correct it.

I gave it some examples from my charts from the blog and told it to take it as an spiration for the next round of improvements.

Here is my prompt:

Too plain and still cluttered. I pasted some templates and examples to look into. Add more white space between the Headline (Main message) and the slide’s core, and between the slide’s core and the bottom. Round the show-ups to whole numbers, since the decimals don’t change the story. Let’s try one more time.

And it came with this:

It still wasn’t lean enough. One big chart made the 42% feel dramatic, but it hid the two things actually driving it: fewer people signing up and a lower share of them showing up once they had.

This version borrowed structure from real consulting slides — a blue headline, one chart, and a labeled column: What, So What, Now What. It’s a legitimate way to organize an argument, but it also leaned on the label ” So What” to do the work the sentence should have done on its own: “So What: Feedback held at 4.6, so quality is not the issue” reads like a form being filled in, not like someone explaining a finding.

I gave it my few final instruction:

Slim down the chart and add the same for sign-ups and show-ups % to showcase the -37% and 96% to 87%. Don’t use blue for the chart; keep it Grey and Orange (let’s not use red). Slim the comments on the right so that another 2 charts fit in. I have made some “space out” changes on the current slide so that you can see how we can get more white in. Match the color of the data labels to the color of the bar itself, always. Can you think about changing the What, So What, and Now What to something more (not so “frameworky”) and leave it black and bold (not blue). And I believe that we need to fix the Key message below the charts. It needs to be a clear CTA on what we suggest based on the story/data. Otherwise, it’s almost there. I would reduce the font size of the data labels by 30%, as they don’t match the overall font size around them.

Before we go to the final output, don’t forget to look at the core Data Storytelling principles, which are described in my simple training for anyone: The Curious Beginner.

Now the final outcome:

I believe the outcome turned out pretty solid. Easy to follow, clean views, key takeaways, and CTA. As I said, the task wasn’t to bring a 100% bulletproof slide, but something you can start to work with, or just fine-tune it.

“Demand fell, not quality” says exactly what “So What:” said, without announcing that it’s following a formula. One Clear Message makes this same point: structure should hold a slide up invisibly, not introduce itself.

The rest of this version’s changes were just as deliberate. One chart became three small ones: show-ups, sign-ups, and the show-up rate.

The palette dropped the template blue for grey and orange pulled straight from the bars, so every label matched what it was labelling. And the closing line stopped observing and started recommending: return to 10 am for two sessions, then compare. Going back to basics, “Start with Why“ is built on the same principle from the other direction — start with the reason, end with the action.

Now what ?

We turned the discipline into something reusable, not a one-off. Every rule that came out of this:

• Two colors maximum
• Labels matched to their bars
• A CTA instead of an observation
• Disclose what you truncate

All of this is now a skill Claude can load before the next slide, so the next brief starts from what this one taught instead of from zero.

Summary – From insights to action

The fastest way to get better at this isn’t to memorize more rule. It’s to experiment with AI on a real slide and keep what works. Every fix in this post came from one round of back-and-forth: a critique, a rebuild, a look at what changed. That loop is now cheap enough to run on every deck you touch, not just the ones that matter most.

Start building your own repository of what “good” looks like. Every time a rebuild lands on something like a color rule, a layout, or a way to close a slide, save it as a reusable pattern instead of a one-off. That’s exactly what turning this session into a skill did: the next slide starts from what this one taught, not from a blank page.

This week, take one slide you already have and run it through the same loop. Ask AI to build it, then critique it the way you’d critique a junior analyst’s draft: too much ink, no accent color, no ask at the end. Used this way, AI isn’t a shortcut around the thinking. It’s a powerful ally that lets you iterate at the speed of your judgment, instead of the speed of your mouse.

Don’t forget to try the free 10 Step Training for Curious Beginner here.

For free resources to help you on your Data Storytelling Journey, click here.

To keep Transforming Your Data Storytelling Journey, sign up for regular Insights.