Ep 642: Most Slept On Claude Feature? Simplest Way To Create Files In An AI Chat

Topics Covered in This Episode:

1. Claude's New AI File Creation Feature
2. Native File Export in Claude Chatbot
3. Claude Pro: Google Drive Sync Capability
4. Python and NodeJS File Creation Method
5. Claude's File Formatting Versus Competitors
6. Enabling Code Execution in Claude Settings
7. Multi-step Agentic Flows for File Creation
8. Top Use Cases for Claude File Creation

Keywords:

Claude file creation, AI file export, create files in chatbot, Anthropic Claude, Claude file creation feature, Generative AI, Natural language file creation, Python and Node.js sandbox, Export files with AI, Google Docs sync, Claude artifacts, AI formatting issues, PDF to PowerPoint with AI, Spreadsheet creation with Claude, Automated data extraction, AI-powered document conversion, Large language model file management, Multi-step agentic workflows, AI data analysis, Automated PowerPoint generation, AI productivity tools, XLS export from AI, Enhanced file formatting, Secure sandbox environment, AI chat prompt engineering, Office Open XML, AI document automation, Prompt injection risks, Memory enabled large language models, Context window limitations, Chatbot native files, Power user workflows, Multi-modal capabilities, AI workflow automation, Research with AI chatbots, Automated insights extraction, AI marketing copy generation, Downloadable AI outputs, AI business use cases, Iterative document updating, Time-saving AI features, Automated bar chart generation, AI for sales data trends, AI-powered slide decks.

Podcast Transcript

Jordan Wilson [00:00:15]:
There's a lot of obvious shortcomings when it comes to today's large language models. Some are serious, like hallucinations and jailbreaking, and some aren't so serious, like when a large language model says you're absolutely right when you give it feedback or how most models can't even create a file. Well, at least that last shortcoming may finally be a problem of the past. That's because Anthropics slipped in a major feature recently into its clawed chatbot, the ability to create common files with natural language. No more copy and pasting. No more getting weird of that weird, getting rid of that weird format. It sounds small, but it's actually kinda big because if you're anything like me, you probably have hoards of helpful chats inside of large language models where the info just sits there. All those golden nuggets that just go unpicked, and unused because you forget to do the copy paste export dance to extract and save those valuable insights.

Jordan Wilson [00:01:20]:
So that's what we're gonna be tackling today. What I think, at least right now, is the simplest way to create files in an AI chatbot and probably Claude's most slept on feature. Alright. Welcome to everyday AI and our AI at work on Wednesday series. Thanks for, tuning in. My name is Jordan Wilson. And if you're new here, this thing, it's for you. It's your daily livestream podcast and free daily newsletter helping everyday business leaders like you and me make sense of all these AI updates that are coming at us nonstop and helps us make use of it and actually save time, grow our companies, and our careers.

Jordan Wilson [00:02:00]:
So if that's what you're trying to do, awesome. It starts here with the live stream podcast unedited, unscripted. But if you wanna take it to the next level, make sure to go to our website at youreverydayai.com. There, we're gonna, recap just the highlights from today's show as well as all of the other AI news that's important. Alright. Let's talk file creation in Claude though. So stick around, And by the end of today's show, you'll understand why traditional AI chatbots really struggle to create simple files. Like it shouldn't be that hard, but for some reason it is, you'll also know how Claude makes this work and the best practices for how to use it.

Jordan Wilson [00:02:43]:
And you're you're gonna get some insights on some of the best use cases for file creation inside Claude. Also stick around to the end. I'm gonna tell you how to get our guide on the 10 best use cases for file creation inside Anthropic Cloud. So this got slept on, and here's the reason why. Anthropic sometimes is one of the most innovative companies, but they're also their their go to market strategy is pretty bad, I'd say. Mainly because some of their better features just get rolled out to their max users, which is those people on the 100 or $200 a month plan. Let's be honest. When Chad GPT or Google do something, have a huge feature, and it's only available to those on the highest tiered plan.

Jordan Wilson [00:03:38]:
It's usually something kind of groundbreaking, right? Like when Sora first came out or, you know, Google, same thing when, their v o three first came out. So when Claude rolls out file creation, yeah, it's really cool, but it's not a sexy feature that's gonna grab headlines. And Anthropic is very far away in terms of number of users from Gemini and Chatt GPT. So that's why some of these helpful features, if they don't get rolled out to everyone at once on Claude, you might miss them. So they actually rolled this out to just the higher tiered people in September. And then about last week, they completed the rollout to everyone. Sorry. All paid users, in Claude.

Jordan Wilson [00:04:25]:
So here's kind of the new capabilities at a glance. So, now I think this solves one of the biggest productivity problems, actually making native files inside of an AI chatbot. Like I said in in kind of the the opening there, you probably, whether you know it or not, you probably have so much incredible information inside of ChattGPT or Gemini or Claude or Copilot that just sits there unused, and it could probably really help you. And one of the reasons why, well, if you've ever done this and you're trying to copy and paste something from a chatbot into a Google doc or a Microsoft word document or an Excel sheet, sometimes Large language models have very strange formatting. It adds these extra tabs sometimes or a strange, background color that it seems like you can't get rid of. Right? So not only is this a huge, leap forward in capabilities, if you are a pro user of Claude on the $20 a month plan, and you can export files, but you can also sync them to Google Docs, which is actually a huge time saver for me personally, cause I usually use between, I think like three computers on an ongoing basis. So even downloading files for me sometimes, I'm like, that's not very helpful because then I'll forget to sync them, you know, in my Google Drive. And this also tackles that copy and paste time trap that I talked about in the formatting wars that plagues many power users.

Jordan Wilson [00:05:58]:
So here's some of the new, changes and some specifics. So like I said, they rolled this out originally in September, but now in October, it's going out to all pro users. Here's how it works. It essentially runs Python and node dot j as a JavaScript file in a sandbox environment behind the scenes. And I'll tell you in a little bit why that's actually extremely important. And it automates the conversion analysis and extraction workflows through simple chat prompts, eliminating the tool switching that you would normally have to do. Right? So I do this constantly in chat, GPT, and Gemini. What I have to do, if I really want something, I either have to copy and paste it in the formatting.

Jordan Wilson [00:06:41]:
Y'all it's it's one of those things. If you know, you know, if you're a power user, you know those little annoying things that it's like, oh, this should take thirty seconds to to fix and reformat, and then it takes ten minutes. Right? Sometimes I literally have a, text edit document open, and I paste it in. I stripped I stripped all formatting. I copy and paste it into a Google Doc and reformat it because, yes, some things are just that difficult, when it comes to this sticky, weird formatting of large language models sometimes. But it's not all perfect fun and games. There are added security risks in doing this, which I'm gonna talk about as well as reliable, reliability issues. There are still some formatting, tweaks that you're gonna have to do and there's, like I said, still limited access.

Jordan Wilson [00:07:29]:
It's only for paid users. Alright. Here's how to enable it. So if you are on any paid account in Claude, you're gonna go into your settings. On the left hand side, you're gonna see a capabilities tab, and then you are going to enable the toggle that says code execution and file creation, and that's it. And you will have to kind of prompt, Claude to do this. Right? Hey. Create a PDF.

Jordan Wilson [00:07:56]:
Create a PowerPoint presentation. Create a spreadsheet in an XLS format. Right? Create a Word doc, dot doc. Right? Sometimes if you're not too specific, it won't do it. It'll just create a normal artifact. Right? Which artifacts in Claude are great. Right? They're, kind of an interactive, window where you can render and run code, but also it will, you know, can save like a document there. But if you don't ask it specifically, sometimes it won't do it.

Jordan Wilson [00:08:27]:
So unlike, you know, certain features in chat GPT or, Gemini, this isn't a toggle. Right? File creation. There's no file creation toggle. You just have to enable it in your settings and then prompt and tell it to create a specific file, and that'll do it. This is kind of technically a preview of, the newer skills feature in, Claude, which I don't know by by the way, like, I don't do a ton of Claude stuff and there's reasons, and I'll probably get to those reasons here in a minute. But if you wanna see me do something on the skills feature, that Claude rolled out about, about a week or two ago, just let me know. Just drop skills in the comments or, you know, if you're on the, listening on the podcast, which we appreciate on Spotify, just type in skills. Alright? So I'll know if if people, really wanna see that or not.

Jordan Wilson [00:09:17]:
So that's how you enable it. Are you still running in circles trying to figure out how to actually grow your business with AI? Maybe your company has been tinkering with large language models for a year or more, but can't really get traction to find ROI on Jenna AI. Hey, this is Jordan Wilson, host of this very podcast. Companies like Adobe, Microsoft, and Nvidia have partnered with us because they trust our expertise in educating the masses around generative AI to get ahead. And some of the most innovative companies in the country hire us to help with their AI strategy and to train hundreds of their employees on how to use Gen AI. So whether you're looking for chat g p t training for thousands or just need help building your front end AI strategy, you can partner with us too, just like some of the biggest companies in the world do. Go to your everydayai.com/partner to get in contact with our team, or you can just click on the partner section of our website. We'll help you stop running in those AI circles and help get your team ahead and build a straight path to ROI on GenAI.

Jordan Wilson [00:10:29]:
Alright. Let's put AI to work y'all. Our new Wednesday series, we do live demos. Nothing ever goes wrong. That's a joke. Doing live demos of generative AI is extremely dangerous. Am I dangerous? It just means it might blow up in my face and, make me look, silly, But let's hope it doesn't, and let's go ahead and get started. So I have two prompts kind of, ready to go, and I'm gonna get this started and, do it live.

Jordan Wilson [00:11:00]:
Alright? So, I have my first one going. That is the exact same one. Alright. Cool. Yep. That's the same one. So good thing I, stopped myself from doing that. Alright.

Jordan Wilson [00:11:16]:
So I started my first one and I'll tell you all, what it is here in a second. Let me get my my second one going so I don't have to copy, type this one out. This one takes a little while. And I'm actually going to, do Opus for this one. Alright. So we're getting a little variety. Alright. So let me tell you, now that it's working live.

Jordan Wilson [00:11:40]:
Alright. Let me tell you what this first, kind of example prompt was. All right. So let's think of the capabilities, of this new file creation. So yes, it can create files, which is cool. Right. But think large language models are multimodal. Right? You can input different file types as well and convert them and mix and match them.

Jordan Wilson [00:12:02]:
I wanted to do something fairly simple, that I knew Claude would be able to handle or hopefully be able to handle. So So in this case, what I did is I uploaded a PDF. Alright. This is one of my shows that I did, a while ago, maybe, I don't know, three to six months ago. I so in the prompt, I said this, this presentation so the presentation is a PDF that I uploaded is about six months old. Research deeply info from July 2025 to October 2025 and add three more slides in the middle of this presentation and save it as a PowerPoint file. Your only deliverable is an updated presentation saved as a PowerPoint. Yeah.

Jordan Wilson [00:12:40]:
I did do this prompt a couple times because the first time even when I said save it as a PowerPoint file, it didn't. Right? So, yeah, you might have to say it two or three times, or if your prompt gets a little long and you say it once in the beginning, you might just have to sandwich it again at the end. There's a technical term for that, but, yeah, it actually works with large language models. If you have something longer, sandwich it, say it once in the beginning, once in the end. Then I said keep the rest of the presentation the same, but you need to find a seamless way to create the updated content and ensure it's styled and formatted the same way it uses the same tone of voice and angle of messaging. Alright. And then I gave it, the link that went along with that podcast. So if you'll find more information, right, because, obviously, I wanna make sure that it goes looks at the entire transcript that went along with this podcast, presentation.

Jordan Wilson [00:13:30]:
Right? So when I say presentation, if you listen on the podcast, I always have a, you know, I'm always sharing my screen. It's usually nothing overly visual. In this case, nothing crazy visual, but I'll explain this to you here in a minute. But I just uploaded a PDF. It's usually the slides that I share on the screen during the podcast. So a couple of things that are hopefully going on that the model will do. It will read through the file I uploaded. It will go and look up additional information on the website that I gave it.

Jordan Wilson [00:14:03]:
Then it'll hopefully go out and do some research. Then it's probably gonna run a lot of Python code and start creating and converting. Right? That's the other thing. I gave it a PDF. I asked for a complete PowerPoint file. So it's going to probably again, generative AI is generative. You're gonna get something maybe a little different over time. It's probably going to recreate this PDF presentation in a PowerPoint file.

Jordan Wilson [00:14:30]:
It's gonna try to make it look as much as it can like the original, which I just created in Canva. It's nothing, like I said, overly visual. And then it's gonna, insert in hopefully a seamless way that updated and fresh information. Alright. So we're gonna look. It's still cooking. We're gonna give it a couple of minutes to cook. Alright.

Jordan Wilson [00:14:53]:
Now, I'm going to jump into the second one that I did. All right. So podcast people, you're not missing anything yet visual, but I will tell you when something happens. So in this prompt, I am saying in August 2025, a 16 z came up with a list called the top, and this list is act actually wrong, or my prompt is wrong, which I do on purpose sometimes. If you watch the show, you know that. Testing large language models, sometimes you wanna throw them a little bit of a curve ball to see if they can still find their way. So I said it's the top 10 gen AI consumer apps, and I believe it was actually the top 50. So, we'll see how Claude does.

Jordan Wilson [00:15:35]:
It should do fine. Oh, and by the way, the first version where it's recreating, I'm using SONNET 4.5. And on this prompt, doing the a, a 16 z list, I'm using Opus 4.1. So what I'm saying, is you will research deeply, right, focusing on the most up to date information. But I said from this list of 50, not 10, your task is to deeply research the top 20 from their list and then deliver me three different things. I want a downloadable, a a downloadable spreadsheet in XLS format with the top 20 on the list, their name, company, headquarter, location, number of employees, CEO's name, money raised so far, last reported revenue, and then you, unique product and features. Just that right there is a ton of research because the a 16 z list does not have all that, obviously. It just has like one or two of those six different data points that I'm asking it.

Jordan Wilson [00:16:37]:
So not only is it gonna have to go find, this list, the correct one, it's gonna have to do a ton of research, run a lot of code to create this, this spreadsheet and, for me to download. Alright. That's deliverable number one. Deliverable number two, I'm saying a downloadable PDF that sorts the top 20 by category of tool or use case and gives short marketing copy copy overview of the company and who can use their tools and for what use cases. Right? So going from, kind of bullet points and spreadsheets to practical, use cases and some marketing language and to sort all of them by category. Alright. And then last but not least, I'm asking for a downloadable PowerPoint that conveys this information. And here I'm giving the model a little leeway.

Jordan Wilson [00:17:27]:
I'm saying you can put it together however you see fit, but make sure it's info packed yet sleek in aesthetics in a clean presentation. Alright. So, if that one's working, We're gonna give both of those some time to cook. So if you're, listening on the podcast, what's essentially happening now is the model is going through multiple steps. Right? So the live stream audience can see it. It's right now the second prompt It's writing, it looks like it's writing some title slides right now in HTML. The other one is still, writing some Python code. So, happening right now in a secure sandbox environment.

Jordan Wilson [00:18:11]:
So let's jump back in and go over the details, and then we're gonna check back in at the end. And we'll see if we cannot make this show drag on just to get the results. Alright. Why do most large language models fail at doing this? Right? I will say this. Copilot, Microsoft Copilot is not a large language model. It is a system that uses other large language models. Mainly it uses OpenAI's models, but, in some of their new agent releases from September, they did start to use some of Claude's, anthropic Claude's models as well. So for the most part, why have traditionally Claude up until now, but also still ChatGPT and Gemini struggled to create files by default.

Jordan Wilson [00:19:05]:
Right? If you use an agent mode, that's a little different. Right, these as an example, ChatGPT agent mode can nice and easily create, Excel sheets, PowerPoint, etcetera. Although they don't usually look that good, and it takes a very long time. But for the most part, large language models are sandbox. So their direct file system rights are intentionally disabled for safety. So it's actually a built in feature that most models, non agentic models by default cannot write files. All right. Yeah.

Jordan Wilson [00:19:39]:
A lot of people don't know that. And they're like, oh, these large language models are so dumb. Originally they weren't built, to do that. Also, there are some complex and proprietary Office open XML specs, that are kind of unforgiving. So it's not easy. Right? It it's really not. That's one thing actually. Microsoft Copilot doesn't get a ton of, doesn't get a ton of fan fanfare, for its ability to create, obviously, you know, PowerPoints and, Word docs and Excel sheets because it's not easy for third parties to do it.

Jordan Wilson [00:20:19]:
And what happens, you like I'd like I've already said, users waste hours. Right? Especially if you are a large language model, heavy team or heavy organization. You probably spend more time formatting, cleaning, reformatting, combining, outputs then it actually takes the model to generate something. Right? So here's how it actually works and why I think Claude's solution is the simplest way to create files in an AI chat and one of the most slept on features. So it builds authentic files by running Python and node JS in isolation. So that way it is a less of a security concern. It kind of isolates that process outside of your normal, chat window, right, without getting too technical. That's kind of how, this works.

Jordan Wilson [00:21:17]:
And then it can generate we went over most of these types, but it can generate, dot doc, XLS, from, for spreadsheets, PowerPoint, PPTX, or PDFs on demand, and then it can also export them to Google Drive. So this bundles the conversion, analysis, and extraction into one conversational multi step workflow, and it reduces, like I said, the app switching. It just allows for faster drafts, fewer handoffs, measurable time savings, and the thing that I like most, you're leaving less of your knowledge and insights on the table because let's be honest, y'all. I would say even for me, right? I'm starting to work Claude. I'm trying to work Claude more into my workflow. Right? But I'm gonna get my bone to pick, which many of you know what's coming. I've been trying to work this in more, mainly because I do so much in-depth detailed work inside large language models, and I probably only end up saving 10% of them. Yes.

Jordan Wilson [00:22:24]:
Large language models have memory, so it can reference those things. But again, I am a power user. Right? A lot of times, I'm doing three, four, eight different versions of something very similar, and I only end with one kind of final version. But if you just have memory enabled or if you can reference past chats, you don't know if it's gonna get one of the first versions that was more iterative or a final version. That's why this is very important. Alright. You know what? I'm gonna go ahead and just get my bone to pick with, Anthropic right now. It took me so long to come up with these demos to do because I wanted to show something impressive, that someone could see and be like, oh, wow, this is cool.

Jordan Wilson [00:23:06]:
This guy did this and you know, the fifteen minutes or ten minutes, you know, in a podcast that's helpful. So many things in a single prompt. I don't know if it's the way that this file creation process tokenizes information. I'm not sure. So the example, the first example, I uploaded a, PDF presentation, an older one, and I just had it, update three slides or add three slides. When I tried to update an entire presentation in one one prompt, one prompt, nothing else. Right. Upload a PDF, gave it very similar instructions as the one I read out loud here, ten minutes ago.

Jordan Wilson [00:23:48]:
The only difference instead of adding three slides with current information, it was just updating all slides with current information. And that did not fit in the context window. Right? Literally, it couldn't complete that. I tried it four times and at the, you know, presumably halfway or two thirds of the way through, it says, oh, you ran out of win, you ran out of room in this context window, start a new chat. So, yes, this feature sounds amazing in theory, especially when you can start, you know, putting together multiple documents. Right? Imagine some of the use cases, which I'm gonna go over here in a couple of minutes, dropping many different file types and then having it do research and all of these things. But in unless you're on that max plan, the utility is actually fairly limited, for that very reason. The context window, again, I don't know if it's if it's how, this this process is tokenized because it's actually not a ton of information.

Jordan Wilson [00:24:46]:
Right? Maybe it's because it's going in loops and having to, you you know, regobble up this information over and over, but it's actually not not super, not super helpful unless I think you're on a higher plan. We'll see if the a 16 z example works this time, which that one, I would say is pretty impressive, but it is more straightforward. It's kind of one round of research and then repurposing those results in different formats. Right? So something a little more iterative, a little more hands on and might not work even if you're on that, you know, dollars 20 a month paid plan. Alright. So let's jump in. It looks like yeah. It looks like we got a couple that are done.

Jordan Wilson [00:25:28]:
Let's take let's take a look. Alright. And here's where I will try to describe to our livestream audience what I'm seeing on screen. So I'm not actually going to download these files because Claude has the artifacts feature, where it's going to preview them. So, essentially, what you're seeing on my screen is more or less what it's ultimately going to look like. So, one thing I'll, I'll tell you. It actually does an okay job at creating PowerPoints. Right? So I uploaded a PDF and it did a pretty decent job of trying to recreate it.

Jordan Wilson [00:26:06]:
It obviously didn't have, you know, the screenshots or, you know, as an example, I've my kind of, my buckshot there on the front, so it didn't have access to some of those things. But aside from that, it did a pretty decent job at rebuilding the old version of the slide deck and I can again, this is a PowerPoint file that I can download or there is this button. So in the upper right hand corner, there's a download button or there's an open in drive. All right. So, I can also slide and toggle to give this kind of more room. So my chat kind of is on the left and the artifact, which is rendering this PowerPoint is on the right. So I'm gonna scroll through. I'm actually gonna scroll down and and see, where it, added, some new files.

Jordan Wilson [00:26:55]:
Okay. So it looks like it added, slide nine, ten, and 11. So, this is a 19 page presentation and it looks like it added slides nine, ten, and 11. So let's just do a quick check. Here we go. So it says, summer twenty twenty five, some new information here, Salesforce, Agent Force, Microsoft going all in with Copilot Studio, Marc Benioff admitting something at Dreamforce conference. All right. The great AI agent failure wave of twenty twenty five.

Jordan Wilson [00:27:29]:
Alright. So, okay. Did did a pretty good job at adding relevant factual, right? I'm looking at these these bullet points. These bullet points are factual, right? Talking here on, page eleven, one of the new pages about AI coding agents, codex, copilot, and cursor. Right. So I don't think codex as an example, was announced yet the first time I did this episode. So this did a pretty good job. Right? So I can, pretty easily, you know, download this, make any updates that I want to, save it to, Google Drive right away.

Jordan Wilson [00:28:08]:
So pretty helpful. So let's go to, that's not the right one. Let's go to our next one. Here we go. Okay. So pretty good job here and it is done. So we have the three different things that we asked for. We asked for a downloadable spreadsheet that gave the top 20 out of 50, as well as the CEO, the company name, company location, number of employees, CEO's name, money raised so far, last reported revenue, and unique product or features.

Jordan Wilson [00:28:46]:
All right. So, I'm showing here on my screen, a spreadsheet here. Let me scroll down. Yep. It looks like it gave me, the top 20. Like I asked, I'm looking at the list. This is correct. I know one through 20 from that list.

Jordan Wilson [00:29:02]:
Cause I've looked at it quite a few times. It got the, the HQ CEO employees funding valuation. So yeah, I'm looking at the valuation. Some look fairly accurate. Like as an example, I think, it in here has open AI's valuation at 300,000,000,000. They did have a recent, this is only about three weeks old. There was a recent story that came out, that put their valuation at 500,000,000,000. So, still, you still need to go through and double check this.

Jordan Wilson [00:29:36]:
If you know these things in your head floating around like me, little easier to, check. But I'm looking at everything else in the spreadsheet. Everything else looks fairly accurate. So I'd say probably went through with maybe 95% accuracy. Right? Which, again, if you hand this off to a junior researcher, it's gonna take them a couple of hours. Right? Pre AI, it's gonna take them a couple of hours and you might hope for, you know, 95% accuracy anyways. Right? That's why everyone's always like, oh, you know, AI doesn't get everything right. Show me show me the human that hasn't made mistakes yet.

Jordan Wilson [00:30:09]:
Alright. So the second thing I asked for, let's scroll down to it. And a lot of work here. Right? Jeez. I'm scrolling through kind of the, the summarized chain of thought here, for cloud, Claude. And I did use Opus 4.1 for this second prompt. A ton. Right? A ton of research that it did.

Jordan Wilson [00:30:32]:
I'm trying to look. It looks like it went to at least maybe, maybe 50 yeah. About fifty, sixty, 70 web pages right there. A lot of steps, a lot of, running this Python code as well to get these final files. Let's just quickly look at the last two files. So, here is our document, our PDF document. Alright. So we have a nice little, cover page here.

Jordan Wilson [00:30:59]:
Nothing nothing special. I didn't ask for anything overly visual in the PDF document. I just asked for them to be categorized. So there we go. We have the general AI assistance and chat bots. Those, from the top 20. Number two, the second category is creative content generation. So you have your mid journeys, Leonardo, Cling in there, Eleven Labs, etcetera.

Jordan Wilson [00:31:26]:
Then you have your developer encoding tools like, Google's AI studio, Cursor Lovable. Then you have your research and productivity tools. Last but not least, entertainment and social AI. So it did a really good job. I I would have to go through and, look at all the fine points, but it looks like it's getting some up to date and accurate information. Right? So for Chatt GBT, it says Chatt GBT remains the undisputed champion of consumer AI with 800,000,000 weekly active users. That's correct, and it's very up to date. Right? I think OpenAI announced that 800,000,000 weekly active users, at their dev date two weeks ago.

Jordan Wilson [00:32:03]:
So it's getting and grabbing very up to date and accurate information. So, looks like test number two did a really good job. Alright. And then number three, the output. Gave it a little bit of leeway. Just said create a PowerPoint. Make it look good. You can choose how to present everything.

Jordan Wilson [00:32:19]:
So let's give it, it's loading here in my browser. Alright. Our cover page, we have a nice, bright blue, color here. The second page nicely formatted with some, kind of breakout boxes highlighting, some certain key facts and figures. So pretty good job, you know, pulling some nice visuals. It's nothing that, you know, I'm blown away by visually, but it's much better, as an example than something that Chad GbT's agent mode would put together. So like I said, Chad GbT's agent mode can put together spreadsheets and PowerPoints. The PowerPoints are, extremely terrible.

Jordan Wilson [00:33:00]:
This PowerPoint is decent. Right? Again, maybe if you hire a college kid or, you know, there's a junior, researcher on your team that's not a, you know, not a design, kind of first person. This is about what you would expect. Alright. So we have a nice bar chart here. Then we have AI categories, who's leading where. Nice, those six boxes with, different, kind of top bar lines. So from a visual standpoint, you know, we got a nice, pie chart here.

Jordan Wilson [00:33:37]:
Did a pretty good job. Right? At least from a visuals, getting the information out there, gave me a quick reference chart here at the very end. So, again, at least this is a passable PowerPoint, right, where a lot of times, if you're using, you know, Chatt GBT's agent mode, not always very good. Or if I'm being honest, the older version of Microsoft's even PowerPoint, with Copilot, unless you're working off of a, template, Even creating it from scratch, I think this is maybe even better or at least on par with how Microsoft does it within PowerPoint using Copilot. I think it's gotten a little bit better over the last few months, but, when I gave it a, a more in-depth look, earlier in 2025, this looks up to par or maybe even a little bit better. Alright. So, there you have it. But let's just quickly wrap this up.

Jordan Wilson [00:34:37]:
It's not without its flaws. Right? There's still, security risks like prompt injections that can allow malicious file instructions to access sensitive data. There's also been reports of frequent downtimes and inconsistent outputs, and there still is some problems with complex formatting and it often fails at requiring extensive manual corrections or recreating documents. But like I said, I think usually, what you get from the file creation out of Claude, I think it's it's it's always, always, always better than anything you would get from an agent mode out of chat g b t. But for the most part, the formatting, I'd say, is clean 95% of the time. Those those hiccups from a design perspective aren't too frequent. Alright. Let's leave you with this.

Jordan Wilson [00:35:28]:
I'm gonna give you a couple example of time saving use cases. So, one is just file conversion. Right? So as an example, uploading a 10 page market research report and converting it to a five slide PowerPoint deck. Right? Probably some a lot of us do. Extracting key points from a lengthy word document into presentation slides automatically or transforming, unstructured PDF data. Right? I used to do this all the time back in the day. Go through all these, you know, RFPs, 10 k's from big companies and, you know, long PDFs and having to grab all this unstructured, data and then putting it into an excel sheet with a single prompt. Another use case, data analysis, that can then create something visual.

Jordan Wilson [00:36:12]:
So in this instance, Claude can ingest raw CSV sales data to detect trends in, compute metrics. Now you can have a marketing manager as an example, generate bar charts even if they don't even really know or understand, what most of the data is. Right? So uploading a CSV, and then exporting or creating a PowerPoint with a bunch of bar charts that you need. Another example, actually, you know what? The rest of the examples, we're just going to give them, I don't want to, take up too much time. But if you do want, more examples, I have 10 actual, very specific, use cases with example prompts as well on, I think, some of the best time saving examples. And a lot of these are stacking, multiple elements, kind of like what I showed you more like, multi step agentic flows, which I think y'all like. I went on a little rant about this last week. Like there's all all these like demos.

Jordan Wilson [00:37:12]:
Right? Everyone's just doing one little piece. Right? They're going in and, oh, here, I'm taking a, you know, my my bullet points and I'm making a PowerPoint. It's like that's not how people work. Right? The way people work is you're doing multiple rounds of research. You have multiple documents, CSVs, all of this, multiple iterations to create a PowerPoint. So in our, example, use cases, many of them are iterative. They are multi step because I think that's how people work. And when you talk about time savings and creating files, this is how most of us are doing it.

Jordan Wilson [00:37:47]:
So if you want access to that, just make sure you find this very LinkedIn live stream URL. So if you're listening live, you know what to do, just click that repost button on LinkedIn. If you are listening on Spotify, check out the show notes. We always leave a link to today's live stream on LinkedIn. Go repost this, and I will share it your way. So I hope this was helpful putting AI to work on Wednesdays. So now, you know, don't have to waste, sometimes hours a week. Number one, you don't have to anymore.

Jordan Wilson [00:38:25]:
You can use this, new setting in cloud, but number two, so much of your best outputs, all of these insights that you probably spend a lot of time on, they don't have to die there in some random chat. Right? That's another time saving element. I spend so much time going into old chats from weeks ago, right, months ago, and being like, what was that? Mainly because I probably was either too lazy, too busy, or I didn't wanna copy and paste and do all this formatting and save it. Right? So now, at least inside Claude, it's a nice way to pull and extract and even if you use Google Drive to sync that knowledge automatically. So I hope the show was helpful. If so, let me know about it. Share, but also go to your everydayai.com. If you haven't already, sign up for the free daily newsletter.

Jordan Wilson [00:39:13]:
We're gonna be recapping the highlights from today's show and giving you all the AI news you need to know to be the smartest person in AI at your company. Thanks for tuning in. Hope to see you back tomorrow and every day for more everyday AI. Thanks, y'all.