Say you need to present last month’s sales results. You have an Excel file, a few notes from the team, and ten minutes on the meeting agenda. You probably don’t need twenty slides. You need to explain what changed, what it means, and what should happen next.
That’s the task I’d give an AI presentation maker. Start with the information you already have, work through the analysis, and then build the deck around the points worth discussing.
Here’s how I’d approach it. I’ll use a fictional online store throughout, so the numbers are easy to follow. This is an illustrative example, not customer data or a product benchmark.
1. Start with the decision, not the slide design
Before uploading anything, write one sentence describing what the meeting needs to accomplish.
For this example: “Review April and May performance and decide what we need to know before increasing next month’s marketing budget.”
That gives the presentation a purpose. A client update might focus on work completed and next steps. An investor presentation needs a different structure, closer to the opportunity, evidence, and funding request covered in HostAdvice’s business-plan guide.
Tell an AI such as Julius, who will be in the room, how much time you have, and what decision is on the table. I would start with six slides for this example and only add more if something important is missing.
2. Give the AI your source data and definitions
Use the spreadsheet or report behind the presentation, not just a prompt asking for “a sales update.”
Our fictional store has these monthly totals:
| Metric | April | May |
| Website sessions | 10,000 | 15,000 |
| Orders | 200 | 225 |
| Revenue | $20,000 | $22,500 |
| Marketing spend | $6,000 | $9,000 |
| Conversion rate: orders / sessions | 2.0% | 1.5% |
Revenue increased 12.5%, while sessions and marketing spend each increased 50%. The conversion rate fell by 0.5 percentage points, or 25% relative to April. Those are different ways of expressing the same decline, not interchangeable numbers.
Before uploading a real file, check the date range, currency, and metric definitions. Say whether revenue includes refunds and whether a blank cell means zero or missing information. Flag duplicate rows for review rather than deleting records just because two orders happen to have the same value.
Only upload information your organization allows you to share with the tool. Remove customer names and other details the presentation doesn’t need.
Keep an unchanged copy of the source file, too. You will need something to check the slides against.
3. Review the analysis before asking for slides
I would separate this into two requests. First, work out what the data supports. Then build the presentation.
Start with a prompt like this:
| Review the attached April and May results for an online store. The audience is the founder and marketing lead. We have ten minutes to discuss whether to increase next month’s marketing budget. First check the totals and calculate the changes in revenue, sessions, orders, marketing spend, and conversion rate. Flag missing information or inconsistencies. Then propose a six-slide outline. Separate findings supported by the data from possible explanations. Don’t generate the presentation yet. |
For our example, a useful outline would cover the decision, revenue growth, traffic and conversion, marketing spend, unanswered questions, and next steps.
The important distinction is between describing a change and explaining it. These figures show that more sessions produced proportionally fewer orders. They don’t establish that the advertising caused the revenue increase or that the checkout experience got worse.
A change in traffic sources, product availability, or tracking could be worth investigating. None should appear as a confirmed explanation without supporting evidence.
Read the outline before moving on. It’s much easier to remove a weak argument here than after you’ve built several slides around it.
4. Generate the PowerPoint, then check it against the source
Julius’s AI presentation maker supports turning Excel or CSV data, documents, and prompts into editable PowerPoint presentations. I work at Julius, so I’m using it as the example here rather than presenting this as an independent comparison.
Once you’re happy with the analysis, upload your brand guidelines or describe the style you need, then give a second instruction:
Create a six-slide PowerPoint using the approved outline and calculations. Use a simple business layout with one main point per slide. Label the reporting period and units on every chart. Keep unverified explanations in an “Open questions” section. Include source references and calculation notes. Export as an editable .pptx file.
Check the first draft against the spreadsheet. In this example, May revenue should still be $22,500, conversion should still be 1.5%, and the presentation should not describe revenue as profit.
Also check what “editable” means in the exported file. You should inspect the text and chart objects, not assume every visual has editable underlying data. An editable presentation is not automatically connected to the original workbook. Microsoft’s guide to inserting Excel data in PowerPoint explains the difference between linked data, embedded content, and pictures.
5. Make each slide easier to understand
Suppose the first draft has a slide called “May Performance” with five metrics, two charts, and a paragraph of commentary.
I would split it. One slide can explain revenue growth. Another can show the conversion decline. Give the second slide a title like “Conversion fell from 2.0% to 1.5% as traffic grew.”
That title tells the reader what matters without inventing a cause. Add a simple comparison chart, label both values, and move the secondary detail into notes or an appendix.
Keep fonts, colors, and spacing consistent. Microsoft’s PowerPoint accessibility guidance recommends readable text, sufficient contrast, descriptive slide titles, alt text, and a logical reading order. Don’t rely on red and green alone to distinguish results.
For a chart, explain the takeaway in text as well. W3C’s guidance on complex images describes how meaningful text alternatives can communicate the information in charts. “Conversion fell from 2.0% in April to 1.5% in May” is more useful than “conversion chart.”
6. Open the exported file before you send it
Download the PowerPoint and open it in the application you’ll actually use. Check slide titles, chart labels, line breaks, notes, and links. Try editing a text box and inspect any chart you expect to update later.
For teams using Google Slides, Google’s PowerPoint migration guide explains how to open a PowerPoint file in Slides. Review that version separately rather than assuming the PowerPoint check covers it.
Run PowerPoint’s Accessibility Checker, then rehearse the presentation out loud. Leave time for questions. If you’re rushing, cut material rather than speaking faster.
My final check would be three questions:
- Can I trace every important number back to the source?
- Have I separated what we know from what we’re guessing?
- Does the audience know what needs to happen next?
For this store, that next step might be reviewing channel-level performance and checking conversion tracking before deciding whether to increase spend. Assign someone to that work and agree on when to revisit the decision.
