How to Actually Present Data Using AI in 3 Moves
Read Time: 9 Minutes
📱 Long-form version of my Instagram short on Top-T (122k views, 4.8k saves). The 60-second cut covers one principle. Below: all three, plus the prompt I paste into Claude or Gemini before any data deck.
Most people don’t know what to highlight for their audience.
Our default is to process everything on the screen. So we put everything on the slide. Eight bars where two would do. Every axis labeled and every data point labeled on top of that. Four colors instead of one. We figure if there’s enough on the canvas, something will land.
I didn’t learn a framework for this either, until about five years ago, from Miro Kazakoff, a business school communication professor at MIT Sloan. This piece expands on what I picked up from Miro and refined over years of presenting data to executives as Head of Business Development at Lucid.
AI can solve this if you know what to ask for. The catch is that no one teaches you how to actually tell AI to construct a data slide. So the default applies in reverse: you ask AI for a chart, you get the same cluttered, four-color, label-everything output. Just generated faster.
The fix is a three-part framework I use every time, in any tool: Layout, Attention, Narrative. Three principles for selecting what to put on the slide and how to walk your audience through it.
Three things every data slide needs to land
Every data slide that actually lands does three things, regardless of who built it. You. Gemini in Google Slides. Claude. Copilot in PowerPoint.
It controls layout. It directs attention. And it wraps the data in a narrative.
When AI does the chart for you, your job changes from “build the visualization” to “tell the tool which of these three things it’s optimizing for at every step.” That’s the new presentation skill. Three principles. Each one a sentence you can hand any AI tool and get back something a senior leader will actually look at.
Principle 1: Layout
Most AI-generated charts default to noisy. They include gridlines, full borders, both axis labels and individual data labels, diagonal text on the x-axis, and bars thinner than the white space between them.
None of that earns its place in front of a busy audience.
Border. Gridlines. Diagonal labels. Both axis numbers and data labels. Bars thinner than the gaps between them.
No border. No gridlines. Horizontal labels. Thick bars. One label per bar instead of two.
Your job is to tell the AI to cut anything that isn’t doing work. Remove the chart border. Drop the gridlines. Pick either a labeled y-axis or labels on each data point. Never both. Horizontal x-axis labels only. Studies show diagonal text reads about 50 percent slower than horizontal. Make the bars thicker than the gaps between them.
Then lock consistency. If you have multiple slides in the deck, your titles all use the same font size, capitalization, and case structure. Your color choices repeat from slide to slide. Your legends sit in the same spot.
Every inconsistency is one more thing the audience has to silently re-parse. Every second they spend re-parsing is a second they’re not spending on your recommendation.
The shortcut for the AI: “Clean styling, no chart border, no gridlines, horizontal axis labels, consistent title format across all slides.”
Principle 2: Attention
A clean slide isn’t enough. Once the noise is gone, you still need to tell the room where to look.
The default move is to leave every bar in its own color and let the eye land wherever it wants. That’s a coin flip. You hand control of the room to chance.
Every bar fights for attention. The eye doesn't know where to land.
Default everything to gray. Bring one element back into color. The eye lands exactly where you want it.
Color is the obvious lever. It’s also the most powerful, because color carries tone. Blue reads as neutral or positive. Red reads as warning or aggression. Gray reads as “background context, don’t focus here.” But color isn’t the only tool. You can use line thickness, intensity, position (bring one bar visually forward), or sparing data labels (label only the peak you want them to remember).
The presentation move is to advance through the slide. Same chart. Different highlights. Each one matched to what you’re saying out loud. Think of the slide as a four-frame story you narrate over one image.
Principle 3: Narrative (Top-T)
Now that the chart is clean and the attention path is mapped, you need to wrap it in narrative. Otherwise it’s still just a chart with highlights.
I use a framework I call Top-T. Four moves. Topic, Orient, Point, Transition. Each move pairs with one moment in the chart. Same slide, different highlight, matched to what you’re saying out loud.
Here’s how it plays on a real example. I run mid-market sales. Last quarter we tested an onboarding redesign with a smaller team. This is the Q3 pipeline review with the leadership team.
Topic. Frame the stake before you describe the slide. The goal, the bet, or the question on the table.
“We committed to $4.8 million in mid-market pipeline this quarter. Here’s where we landed.”
Q3 Pipeline by Segment
Target $14.1M total · Actual $13.9M total
Now there’s a question hanging in the room. Did we hit it?
Orient. Tell them what they’re looking at, so they stop searching.
“You’re looking at pipeline by segment. The two blue bars are mid-market and enterprise. The gray bars are SMB and partner.”
Q3 Pipeline by Segment
Target $14.1M total · Actual $13.9M total
Three seconds. Everyone knows where to look.
Point. One number. One insight. That’s what they’ll walk out remembering.
“Mid-market hit $5.1 million. The entire beat came from the onboarding redesign team.”
Q3 Pipeline by Segment
Target $14.1M total · Actual $13.9M total
Don’t explain every bar. Pick the one that matters.
Transition. Connect the data to a decision or a question.
“Mid-market is our fastest-moving lever right now. The question is whether we replicate that onboarding fix in enterprise next quarter.”
Q3 Pipeline by Segment
Target $14.1M total · Actual $13.9M total
You just turned a chart into a business decision in four moves. Same slide, four moments of focus.
This is what we teach inside the Impromptu Speakers Academy: frameworks like Top-T that let you speak articulately on the spot without scripting every word. Once you’ve used Top-T a few times, you can apply it the moment someone hands you a slide you’ve never seen before.
Three of those four moves (Topic, Point, Transition) live outside the slide. They live in your voice. The slide just supports them. That’s why a clean chart with focused attention beats a polished slide with no narrative every time. The audience is listening for the structure.
The prompt I paste into any AI tool
Here’s the prompt I paste into Claude, Gemini, or Copilot before I build any data slide. It works the same in all three because the principles are tool-agnostic.
You are designing one data slide for an audience that has roughly
four seconds to decide if the slide is worth their time.
Before you generate anything, check whether I've given you all of:
- Business context (one or two sentences on the situation)
- Raw data (a table, CSV, or clear description)
- The one decision I want this slide to drive
- The one takeaway I want them to remember
If any of those are missing or unclear, STOP and ask me for them.
Don't guess. Don't fabricate numbers. Once I respond, continue.
Once you have everything, apply three principles:
1) LAYOUT. Strip anything that isn't doing work. No chart border,
no gridlines, horizontal axis labels only, bars thicker than the gaps.
Pick either a labeled axis OR data labels, never both. Keep titles,
colors, and legend position consistent across slides.
2) ATTENTION. Default everything to gray. Bring exactly ONE element
into full color or full opacity, the element that supports the point.
If I need to make multiple points on the same chart, give me a
sequence of versions where each version highlights a different
element, matched to what I'd say out loud.
3) NARRATIVE. Wrap the chart in four bullet moves I can speak to in
my own words. Each bullet captures the gist of what I'd say at
that beat:
- TOPIC: the stake (the goal, the bet, the question being evaluated)
- ORIENT: what the audience is looking at on the chart
- POINT: the one number and the one insight worth landing
- TRANSITION: the decision or next question the data points to
Output:
- A chart recommendation (type + which fields map to which axes)
- The four-bullet narrative (just the gist of each beat)
- A list of which element to highlight at each moment of the narrativePaste your numbers, fill the brackets, hit return. The output is a slide spec you can hand to your design tool or build yourself in five minutes.
What changes when AI does the chart
When AI handles the chart, the bottleneck moves. The slowest step used to be wrestling PowerPoint into producing something clean. The slowest step now is figuring out what you actually want the slide to do. A better problem. It also exposes how many decks were getting away with messy charts because nobody had time to fix them.
Tool fluency means knowing what to ask for. Tell the AI to apply Layout, Attention, and Narrative. Pass it the business context and the decision you’re driving. The speaking still lives in your voice. Topic. Point. Transition. AI just clears the desk so you have room to deliver them.
Whenever you’re ready, here are some (free) resources you can check out:
- 7-Day Speaking Course: a week of short lessons that fix the most common patterns I see in client work.
- How To Get Your Message Across Busy Executives: the companion piece on framing for senior audiences.