Why I Stopped Using PowerPoint (And What I Built Instead)

I have worked with over hundreds of software applications in the past two decades, most of them are forgettable while others are the reason I work in digital enablement. But if you know me, you already know my PowerPoint opinion.

I dislike it. A lot.

That's not me being rebellious, it's a calculated position. The amount of effort it takes to build a good deck versus the result is rarely worth it. Yes, there are pitch decks that net millions from investors. But the average corporate worker is grinding through hours of slide work every year for a presentation that people will half-watch while checking their email. At some point, the math stops working.

For years, I had a goal sitting in the back of my mind: get AI to build presentations for me.

Not from scratch, necessarily, but if I came up with an outline, I wanted AI to do the construction. What I found out the hard way is that this is one of those problems that looks simple from the outside and reveals itself layer by layer as you go deeper.

The Obvious Approach (And Why It Failed)

My first attempt followed the logical recipe. I built a Claude Skill and loaded it with our corporate brand guide and PPT template. Set the guardrails, defined the output, tested it.

The results were technically correct. Headings where they should be. Colors on brand. Text in the right places. But they were not presentations I would show my manager or the executive team.

This is a gap that doesn't get talked about enough in AI conversations. There is a meaningful difference between a result that is correct and a result that is presentable. AI is excellent at linear execution. Give it a clear structure and it will follow it. Yet building a slide that actually looks good requires something else, a kind of subjective design judgment that comes from years of looking at what works and what doesn't. AI doesn't have that instinct. The first version of my skill could execute, but it couldn't feel.

I kept tweaking and the results got marginally better. There was just something missing.

The Iteration That Got Me Closer (But Not There)

Around the same time, I was building a separate brand skill for a different problem. I produce a lot of monthly and quarterly reports, and I wanted them to look sharp without spending hours on formatting. The brand skill I built for this was genuinely good. It ingests plain text, tables, and data analysis, and outputs a fully branded Word document. Correct fonts, heading hierarchy, colors, tables that actually look clean. It works because Word documents have one major advantage over slides: they can just go to the next page. There is less spatial constraint, which means less design judgment required.

That skill worked. The PPT skill still didn't.

I pushed the PPT skill further. Built a dedicated version specifically for presentations, trained more tightly on our design standards. It got to roughly 70%. Closer, but still falling flat aesthetically. The design was off. Slides felt busy or unbalanced in ways I could identify but struggled to fix through prompt iteration. I was pushing against a ceiling.

The Reframe

I had a big presentation coming up. It was a Friday night and I was stuck. The deck wasn't flowing and I couldn't get it where I wanted it.

I picked it back up Sunday afternoon and asked myself a different question: if I could present with no constraints at all, what would that actually look like?

That question unlocked something. I spent years doing design work, starting in my late twenties. I rebranded a national organization. I designed logos that went on the sides of buildings in Houston. Billboards to business cards and everything in between. The designer in me has always been at odds with PowerPoint because PowerPoint enforces a set of limitations that good design naturally pushes against.

So instead of continuing to improve the skill, I built the presentation differently. I worked with Claude to construct it as an HTML artifact. Full 16:9 layout. Clean design with real control over every element. And then my systems brain kicked in and I started overcomplicating the hosting question until I hit the most obvious answer possible: just double-click the file and open it in a browser. Done.

That decision unlocked something I hadn't fully anticipated. Most of the presentations I give include live demos. AI tools, data dashboards, platform walkthroughs. In PowerPoint, jumping from presenter mode to a live demo is genuinely awkward. You're fumbling between modes, sharing a tab, losing your place. With the HTML approach, I just have my presentation in one browser tab and my demo in another. Tab switch. No friction.

After a few tweaks in VSCode, which was fast because I know HTML, the presentation was exactly what I had envisioned. I showed it to a few people before the meeting. The feedback was the best I had received on any deck in my current role.

Then Building the Skill Was Easy

Once the HTML artifact existed, I had something concrete to work from. I asked Claude to take that artifact and combine it with the brand skill I had already built to create a new hybrid skill. The logic was straightforward: stay on brand, stay within 16:9, build in HTML.

The skill delivers a presentation file I can open in any browser and present from directly. Because it outputs a bounded 16:9 layout, I also get a clean PDF version as a byproduct, which I use as a pre-read before meetings or a follow-up document after.

This is the 10-80-10 in practice. The first iteration required real front-loaded work, knowing HTML, having a design foundation, building the brand skill separately, constructing the original artifact by hand. That was the 10%. Now the skill handles the 80%, and I do a final pass to make sure everything is right before I present.

The total time savings on an average presentation is somewhere between 70-80% compared to building in PowerPoint.

The Soft ROI

Most AI ROI conversations stay in a few lanes: attributable revenue, time saved, and cost reductions. Those matter, and I track them. But in corporate environments, there is a soft category that rarely shows up in dashboards.

Getting people to say yes is valuable. Alignment is valuable. A presentation that earns attention, communicates clearly, and makes stakeholders confident in a direction produces real business outcomes. It just does not have a clean metric attached to it.

I now have a tool that helps my presentations stand out. The feedback loop is real and meetings land differently. That matters for my team, for my projects, and for how AI gets perceived and adopted across the organization. A sharp output builds credibility for the process behind it.

What This Taught Me

The lesson here is not "use HTML instead of PowerPoint." Most people reading this are not going to do that, and that is fine.

The lesson is about where solutions actually come from.

I followed the obvious path first, the recipe approach, and it took me most of the way there. The final gap required me to stop optimizing the existing approach and ask a different question entirely. The answer came from design experience I built over a decade before I ever touched an AI tool. AI did not have that knowledge. I did. The skill I eventually built is only as good as the expertise I brought to it.

AI executes the process. You still have to know what good looks like.

That is the part people skip when they talk about AI replacing expertise. The tool is powerful. The judgment that shapes the tool still has to come from somewhere.