AI Presentation Tools & Google Gemini Omni

Have you tried any of the AI presentation creation tools?

They’ve certainly come a long way. I remember when they were first coming out, the tools were... garbage (a lot of them still are). They would all make slides that were essentially images that you could not edit and would have to rebuild them through prompting and hope you didn’t get into a doom loop.

I Tried Every AI PowerPoint Tool and They All Failed

I had a somewhat simple task last week to create a slide deck for a customer presentation. Most of the content was already developed, just needed to be merged into the right branding and into a master slide template. That's it.

So I figured I'd let AI handle it.

First stop: the Claude Code PowerPoint skill. It choked immediately. The slide master was too complex for the tool to parse. Dead on arrival.

Fine. Plan B. I spliced the deck together manually and pointed Microsoft Copilot at it from inside PowerPoint. "Take slides 8, 9, and 10 and apply the slide master from the rest of the deck." Copilot cheerfully agreed and then churned through 27 steps over 20 minutes. Then it told me it couldn't complete the request.

Plan C. The standalone Copilot tool, which is supposed to be more capable. Except I had to upload the deck to SharePoint first because standalone Copilot can't touch local files. I gave it the same instructions and walked away. I came back ten minutes later to find it had built an entirely new slide deck with random branding. My content was there, but it didn’t do at all what I wanted. I then asked Copilot why it made a new deck and it of course apologized for its confusion and was ready to do it right this time.

I let Copilot go ahead and try and fix it correctly and it worked for another 10 or 15 minutes and came back and said it was done and ready to review. There was no link in sight, so I asked Copilot to share the link to the presentation in SharePoint. It would provide a link, tell me to click the link, but there was no link?? Ugh. It then decided it would make the link clearer and then worked for a minute or two and nothing. When I told Copilot the link still wasn’t there, it suggested that I shold have the file zipped and then the link for a zipped file would be more stable. This do work, but the deck was still a hot mess that needed a lot of work.

At that point I was about to bang my head on my desk, so I just did the branding merge manually. I just needed to update 3 slides and it took me about 2 minutes manually after wasting at least 2 hours trying to get AI to do it for me.

I’m sure this sounds familiar to you if you’ve been playing around with AI tools, and I apologize if this was triggering! I could have done the whole job in 15 minutes if I'd never touched an AI tool.

This is a problem. Every AI company is racing to be the interface layer for your productivity apps. Microsoft, Anthropic, Google, they all demo these beautiful workflows where you ask the AI to edit your slides and magic happens. The ads make it look incredible. The reality is that the moment your files have any real-world complexity, a non-standard template, merged branding, a complicated slide master, these tools crumble.

The underlying issue is that software like PowerPoint wasn't built to be operated by AI agents. The tools haven't invested enough in the unglamorous plumbing required to make this work reliably.

So, if you're evaluating AI productivity tools for your org, my advice is simple. Test them rigorously on your messy files. And let me know if you’ve tried out any tools that have actually done a great job with more complex and messy PowerPoints.

Google Introduces Gemini Omni

Google announced Gemini Omni at Google I/O 2026 this week. It's the next step in their natively multimodal Gemini line, and the first model focused specifically on video generation and editing.

Omni is supposed to combine images, audio, video, and text as input, and Omni produces high-quality video grounded in Gemini's world knowledge. You edit through natural language conversation, with each instruction building on the last. Characters stay consistent, physics hold up, and the scene remembers prior context.

The first model shipping is Gemini Omni Flash, rolling out to Google AI Plus, Pro, and Ultra subscribers through the Gemini app and Google Flow. YouTube Shorts and YouTube Create App users get it at no cost starting this week. Developer and enterprise API access is coming in the next few weeks.

A few capabilities caught my eye. Conversational video editing lets you iteratively refine clips without losing the thread. You can tell it something like, "when the person touches the mirror, make the glass shatter." Multi-turn edits build on each other. Omni also demonstrates improved physics simulation, handling gravity, kinetic energy, and fluid dynamics more realistically than previous models. The model can create videos from any combination of input types, using reference images, existing video, audio (voice input to start), and text together.

Google is also introducing Avatars, which let you create a digital version of yourself for video content. All Omni-generated videos carry an imperceptible SynthID digital watermark for verification – one way we’re seeing companies combat deepfakes.

Google is positioning Gemini as the creation platform. By launching free on YouTube Shorts, they're betting that democratizing video generation will pull creators deeper into the Google ecosystem. That's a very different competitive play than what we're seeing from OpenAI, AWS, or Anthropic.

Now, just like the AI presentation tools, the demos look sharp but will Omni hold up to the test? I guess we will find out!

My Final Thoughts

The gap between AI demos and AI reality is the push and pull tension we’re constantly seeing.

The hard work now is making AI reliable in messy, real-world conditions. The models are smart enough and the infrastructure is scaling. What's missing is the boring, unglamorous engineering that turns impressive demos into tools people trust.

If you're thinking about where your organization should trust AI, the build vs. buy dilemma, or you have some ideas on AI projects for your organization, reach out.

Feel free to follow this link.

Until next time,
Ryan

Now time for this week’s AI-generated image and the prompt I used to create it.

Create an image of a muppet at a memorial day party. He is wearing a typical memorial day outfit. There is a pool in the background and his muppet friends are playing volleyball in the pool. There is music blasting, the sun is shining, and the muppets are having a cookout.

ok now a big dinosaur crashes the party