AI Tool Playground

My name is Janette Roush and I am the SVP of Innovation and the Chief AI Officer for Brand USA, which is the destination marketing organization for the United States.

And today I am very excited, to share with you AI tool playground. So I'm going to walk through both kind of the tech stack that we use for AI here at Brand USA -different tools that I use personally, tools, in some cases that I don't use that much personally, but I know are important for other folks.

And give you a well-rounded look at what's available for text generation, images, videos, creating presentations, and a little bit more.

And as always, I want to begin by sharing with you my AI agenda for the industry. So first this image, we will be introduced to this tool later on in the presentation, but it is from PIKA Art, and all it does is make cute little videos like this, little animated gifs that I think are really good at drawing in someone's attention.

But then looking more holistically at what is my AI agenda at Brand USA, it's to establish operational, excellence for Brand USA as a destination marketing organization. So making sure that we're taking advantage of opportunities to level up the work that we do internally. Second is industry empowerment.

So how do we take what we are learning internally and sharing that with the broader industry to help all of you in your work? And then finally, the third part is the traveler experience enhancement. So what are we able to do in order to make the United States more discoverable and more bookable by, leisure and business travelers across the world?

So that is my goal and part of the Agents of Change webinar series is really focused on number two, industry empowerment. So I'm going to begin this webinar by talking about frontier models. These are the basis of our work with AI tools. So this would be all of the GPTs that are inside of ChatGPT.

So whether it is GPT-4 oh. GPT oh three as in this example here, all of that inside of ChatGPT is considered, those are considered foundation models for ai. And so I believe the best way to learn how to use AI is to pick one of the frontier models that I'm going to show to you and just invest in a paid version of that tool and invest in the time spent becoming accustomed to what it's good at and not good at.

Thinking about this both from a text perspective, but also from a multimodal perspective, because we're going to go a little bit deeper into image generation and video generation, sections later on in this webinar. Right now thinking about frontier models, this is really the home of where all of the power of working with these tools is going to be.

All of the other, tools I'll show you later, will be built on top of these models. And if you are looking for one model to use at your DMO or your organization, honestly, I think ChatGPT is still the way to go. I would look obviously at a paid account, either a team account if you have fewer than, 150 employees or an enterprise account if you have 150 employees or more.

And just in terms of capabilities, I believe that ChatGPT has the widest, most broadly useful capabilities for most people and for most teams. An ancillary to this that I did not call out separately in this presentation would be using a co-pilot from Microsoft. That tool is built largely on the open AI models.

What makes it different? My understanding is that it's less powerful than working directly with a tool like ChatGPT. The trade-off is that it is substantially more secure, so it's going to have the same security protocols in place largely that you would have with all of your other Microsoft tools.

So all of the work that we do, for most of us that work lives in the cloud anyway, so it does make sense to use services that are already connected to our data in the cloud. If you're a Microsoft shop, that's going to look like copilot. The next Frontier Model tool that I really think is terrific is from a company called Anthropic, and this tool is called Claude.

What I do really like about it though is that it is a terrific thought partner and it is a better writer than ChatGPT.

So whereas ChatGPT, everything I get out of it lately, it's not only filled with M dashes, but it's filled with the sentence construction of, it's not about X, it's actually about Y. And I see this all over LinkedIn and all over the internet. I don't use AI a ton for writing anyway, but certainly if I were to, I'm going to lean a little more on Claude than on ChatGPT.

I also like Claude because if you are having a long conversation, so whether you are sorting out the best strategy for something that you're working on, trying to learn how to do something, anything that would create a really long thread ChatGPT I find gets a little circular in that thread where after a certain number of interactions or when it's context window, which is its short-term memory, when that gets too overloaded, ChatGPT will start to just give you the same responses over and over again.

Claude I find is less likely to do that. It's also just a good way to learn that different models will respond differently to the same. So if you have the ability to subscribe to two paid models, I think that is a great way to learn how AI behaves differently, when it's trained differently.

And Claude is a great option for a second model. The other option, which I think would be good for that second model is Gemini from Google. I do have a personal account that I will use for personal projects on occasion, just to keep up on how Gemini works and how useful that tool is.

I'll say if you are in the Google ecosystem, if you use Google Workspace, you very likely already have access to the paid version of Gemini. And if you are already trusting Google with all of the rest of your cloud infrastructure. I don't think it's, a leap to then also trust Google with your AI infrastructure.

So Gemini, it has less than a of a personality, I think, than either ChatGPT or Claude. So I don't always enjoy the back and forth as much, but this new model, the 2.5 Pro model is on par, with the best models available from both Anthropic and open ai. So it's overall, it hasn't always been a great product.

I'd say today it is. It's a great product. It just doesn't happen to be the one that I use most of the time. And the cool thing about Google is that when you work with it, you also get access to other tools in Google's ecosystem. And one of those that I really like is called NotebookLM.

And this is great if you are doing work that needs to be very grounded in truth. When you don't want AI to hallucinate. So Notebook LM was created by Google researchers who were creating a product that would be helpful for students and for researchers. So this was really designed for an academic environment, and what it's meant to do is make it easier to learn something.

So here you can see that I'm in a notebook. The notebook's name is AI's Disruption of Travel Search and Booking. And here you can see that I can add sources to my notebook. And I have in my NotebookLM folder a number of different notebooks on different topics. So that's a great, it helps you keep your information organized.

But then when you add sources, you are now able to come over to this box, this window. And you can start typing into this window and it will answer your question. And if you click on one of the footnotes that it gives you, it's going to open up your source and it is going to highlight the portion of your source where it got that information that it is using to answer your question.

If you're writing something that has to be factual, this allows you to fact check against your own sources in real time, which makes it a very powerful tool. It also has a lot of capacities to help you learn things. So over here, you might remember this tool from when it became famous in October of last year because it released the opportunity to create a podcast from those documents that you uploaded.

It's called Deep Dive. It has, you see here, two hosts by default, but you can also customize that podcast and so it reads through all of your sources, and then it will create a 10 to 15 minute podcast that sounds like it was ripped from NPR in order to teach. Whatever it is that you wanted to learn from those sources.

Take your last annual report and see what that sounds like when it's two NPR podcasters having a conversation about it. You are also able now to jump in and join the podcasters in their conversation so you can actually ask questions. And then down here you see you can also make a study guide with the materials, a briefing doc, FAQs timeline, or making a mind map.

This is an example of what the mind map looks like for this small notebook that I was putting together. So we're looking at the future of DMOs in a changing world, what are our key challenges? And here you say, oh, it's shifting consumer behavior and what is the role of the DMO? And you can double click on that and it keeps going deeper and deeper into the topics that you're interested in.

Now I'm going to take a couple of minutes to go through some image generation tools. The first one of these is ChatGPT. And so actually I am having, for the ways that I use image generation, which honestly it's either for LinkedIn or it's to punch up a presentation or it's to make a joke in a work email, right?

At Brand USA, we don't want to use AI generated images to promote a destination. But that doesn't mean that destinations and travel companies can't find lots of ways to use AI tools. So here I'm drowning in paperclips. If you are familiar with the paperclip maximizer thought experiment about AI agents, this was meant to illustrate that point.

This is, I clicked on a tab in my ChatGPT account that's called library. And if you click on that tab yourself, it will bring up all of the images that you have used ChatGPT to make. This image right here, this sad face, was particularly interesting because I asked ChatGPT to create an image of what it felt like to have a conversation with me and that was the output.

So that it's a real feel good moment. If you want to see, what does it actually take to create an image in ChatGPT. This I have edited and sped up quite a bit, because these image generation tools are not fast. So in all of these cases, it's taking probably two to three minutes every time I'm making an image.

So here I uploaded my own headshot and I said, please, use this to make an image of me making an image. And instead, that's not me. I wanted this person to be the one making the image. And so I am just using plain language down here to say I actually wanted to show me in the pink shirt using the computer to make an image in ChatGPT make it photo realistic because here you can see they made it an illustration and I wanted it in a horizontal aspect ratio. And so now it doesn't give me any notes back. It just starts to work. You can see it's made a really unsettling, creepy version of that image of me.

Another image generation tool that is very well known and popular is Midjourney. I think a great tip for using Midjourney is that when you log in, you are taken to a homepage that shows you images and videos that other people have created, and that is a good way to learn how to best prompt these tools.

A great way to learn more about working in Midjourney is just to click on one of those images and see what prompt the user used in order to create that picture. Because AI rewards, people who are subject matter experts. So when it comes to creating a campaign strategy, that's something I did before AI existed. So like I'm a super user when it comes to doing that in AI. I could not create images before AI existed. So I am not a super user by any means of the image generation abilities of AI, but for people who are good at creating images, midjourney is particularly unique because it wants you to use the language of images and film when you are talking to it.

And because I don't really have that vocabulary, I can cheat, I can ask ChatGPT to write prompts for me in that style. But it's not, language that comes to me naturally. So I have heard, podcasts where people talk about, they were creating a brand identity for their, solopreneur company.

You can do that by just scrolling through Midjourney and pulling out the images that really strike you as interesting. Then use those images and those prompts to learn more about, okay, what is the style that I like? How do people describe the style that I like? How do they talk about it to Midjourney to create more examples of that style.

We have a question in the chat about whether there are copyright issues. Copyright is a rapidly evolving space when it comes to the US legal landscape.

Originally what the copyright office said is that you must be a human being to have a copyright on an image. My understanding is that if AI makes it, it can't receive a copyright, which means you can't be in copyright violation supposedly by using that.

I know that this is a concern if you are using AI to write for you, AI can plagiarize without knowing it because it's just putting words in a certain order to make you happy. So if it makes up words in a certain order, but they happen to also exist written in that certain order, you would be violating somebody's copyright by using those words in that order. We have a resource here, copyright.com. Julia is sharing as a resource, so that will be good to look into. If AI made an image of an Amazon Alexa device, I don't know if it means, oh, great, you can use that without any kind of infringement because AI made it, because AI is making an image of something that is protected, my understanding is nobody else can have a claim to it because it's the AI that generated it.

And now here we have a screen grab of me creating an image inside of Midjourney. So here you can see you type what your prompt is into this box that says, what will you imagine? And this is very sped up. So this whole process was probably three or four minutes for me to get to this point that you're seeing. And anytime that you generate an image, you see over here, it gives you four different options. Then you can select one of the options and then it brings up opportunities to make edits to that particular image.

You can ask it to vary it, to upscale it, to create it in a different format. None of these images that Midjourney is creating for me here are particularly wowing me.

But also 20 minutes later I was like, okay, I don't actually have a need for this image, and so I moved on with my life. It's an interesting example of what's actually possible with these tools. And I've definitely used Midjourney to create a lot of the images that I share in presentations.

Moving on to the next section, video generation and editing. This is going to walk through what editing in Descript looks like while I talk to you. Descript is an editing tool for video and for audio. I first learned about it listening to podcasts where the podcast hosts were using Descript to edit their podcast and then talking about that on the podcast.

And it's great because you can see over here with all of this text. This is the very first Agents of Change webinar that I did, back in June. And so you are able to work over here. This button for the AI Underlord will bring up an AI video editing assistant that I can say remove all of my, ums and ahs.

And then over here you can see where it deletes out. And then I can hit accept in order to accept those deletions, I can make an AI version of my voice. So if there's something in the webinar that I didn't say the way I wanted to say it, I can actually delete that section from the text type over it, and then it will use the AI generated version of my voice to update the recording.

And all of these recordings are available on Brand USA's website. And then the other thing you're able to do with these tools is, like I plugged this, 101 video. I am in the process of turning that into TikTok educational content.

So if you wanted to have a 101 course that you didn't want to sit down for 50 minutes and just watch it sitting at your computer, maybe you want to watch it on your phone scrolling through TikTok videos. I was able to replace myself with a yarn avatar of myself talking against the background of these slides, using my voice, the actual recording from the webinar, with the hope that it's another way of reusing content.

And I think that is a big push for live events in the near term future, is making sure that when we do a live event, we don't leave all of the great education that happened on that event on the floor, when you come home from DI you have the opportunity to expand on that content.

So you can take those webinar recordings and your AI can go through your entire transcript and then pull out what it considers to be the five most interesting segments. And so it will actually create, social media, promo videos for you.

Do you want it to be oriented for LinkedIn or for Instagram reel, or do you want it to be formatted for TikTok? So it's been a really great tool. They're making improvements to it all the time, and it doesn't require any AI expertise if there is such a thing to use. You are just having a conversation about what you want your video to look like.

I think that's the ultimate promise of these AI tools.

This is another video tool that is called Runway and very similar to all of these tools. It's just text-based. You go into the tool and type what you are looking for. You can offer it images to use as a basis, and then it can create things like a gif, which is what you see here. The prompt that I had given it was that I'm at IPW and I'm going back into my hotel room after a long day and I open up the closet and all of the IPW branded tote bags fall from the toilet and crush me.

So I gave it the IPW logo. As you can see, it doesn't look like that at all, and I don't exactly look like the same person emerging from the closet as I did going in. But again, for fun, simple video editing, absolutely an option.

And then this shows you what the tool looks like when you're using it. Again, I've really sped it up and here I've given it a copy of my headshot and I said, show this woman soaring through the New York City skyline. I was very disappointed with the outcome because why would that be at all what I had asked for, right?

I'd accidentally had done that same prompt twice in a row. So I'm like, maybe the second one is better, eh, not really. But again, these tools reward expertise. Oh, and I have a question from Yolanda. I did not record myself going into the closet, so this entirely AI generated. I just uploaded my headshot and the logo for IPW and told it what I wanted.

And then this I brought up at the beginning of the webinar. There are all kinds of these, small animations that are possible inside of pika. Now I'm going to talk a little bit about vibe coding tools. So this one I am a big fan of. It is called Lovable, and you can find it@Lovable.dev DEV. And in this example, you can see I asked it to create a simple app that helps social media managers create short form content.

But what Lovable is best at is as a front end tool. So making something look really beautiful. So here I use this at IPW to create this, AI walked me through the entire process of both going to Lovable.dev building this.

I then, it told me, oh, if you want it to look a little nicer, we could have an image at the top. This is the type of image that you can create. And I then went to an image generation tool to spin, just a little header for this. this is obviously a screenshot, but you could select the different days of the week that you wanted to make an appointment.

And then when you clicked on this, it would bring up a registration form and then in the backend it connects to a database tool called Supabase, where it allowed me to see who had actually registered for each of those sessions. And so for just making small personal use apps, Lovable is very helpful.

If you also wanted to make a small website for some reason, Lovable is going to be a great tool for that. Sarah just asked, which of these tools are free versus requiring payment? I believe that if you want to make images inside of ChatGPT, I don't think that is available on the free version. I think you need to have a subscription for that.

I know you need to have some kind of minimum subscription for midjourney. I think that is the case for runway as well. PIKA is they do have a free tier available. So I think it's generally, it is more difficult to find a free tier for the video and the image tools just because those are more expensive to run as a product.

But then Lovable, I did this, I still don't have a paid subscription to Lovable, so this entire process was free. and then I made sure that for the backend, which again was this database product called Subbase, I also used a free version of that. So I made sure I didn't collect any PII when I was doing signups.

Like I didn't collect email addresses because I was not fully clear on what the security was of that signup system. agent.ai is another tool that is free to use, for certain tiers. Obviously all of these tools want to get you into a paid ecosystem, eventually. agent.ai is your gateway into creating your own AI agents.

And so this is an area I'm very interested in moving into. an agent.ai is created by one of the co-founders of HubSpot, so this is. Based on years of his relationships. And you could see here we are making an assistant for LinkedIn to hold the profile that you give it of a business event planner and then give you, essentially, yeah, what's the prompt?

You are the very best convention sales person in the world, and pull out all of the information from this profile. That would help me have a really great opener con of a conversation with this meeting planner. And so here you could see where all of the output is from me testing this particular prompt.

And this is the process of creating your own AI agent. You're going into actions and then you are just setting up here. It's just five actions that you are outlining in order to create your own AI agent that pulls from LinkedIn's API the information that you're looking for. And because this is through agent.ai, isn't a fly by night operation because it's through this person who's connected to HubSpot.

He knows the LinkedIn people, so it actually has API access to LinkedIn. So if your prospect has a closed profile, this would not work on LinkedIn. But for anybody who has an open profile, this is legitimately accessing their profile in order to enrich emails that you can send out in the future.

Then this, I is just starting to learn a little bit about this over the weekend. N8N is another kind of vibe coding agent building tool. And so the way this works, whereas an agent.ai, you saw that you outline individual steps. This does the same thing, but it visualizes it as you can see in a different way.

this is a visualization of how you could use N8N in order to create an onboarding process for your IT team. let's say that somebody triggers this workflow or this automation by doing a create new user form submission. That's then going to this tools AI agent. One of the tasks that we'll do is ask, is this person a manager?

Because they're going to be automatically set up in Slack differently if they are a manager than if they aren't a manager. And so this is a good way to start picturing what should AI workflows look like for you, your department, your company over the next year, two years, as this becomes something we all need to start learning how to do.

And then we have meeting and presentation tools. I am building this presentation in beautiful ai and here this is just a screenshot, very meta of what it looked like for me to put together this presentation on AI using beautiful ai.

I just, find it, honestly, it was a terrible learning curve, and now that I am familiar with it, I would have a hard time going back to either PowerPoint or Google Slides to put together presentations. it makes it very easy to put in a grid format. What are my images and texts going to look like?

It's very easy to not only add in screenshots or videos, but to also find other slides from other presentations and add those in. And so that's something I always find to be a pain in other presentation tools. And for me, beautiful.ai has made that absolutely dead simple.

If you have beautiful.ai across multiple users, you can create templates and you can also on those templates, if you were to update your logo. What's cool is that logo will not only update in one spot, but if other users use that same template, that new logo will populate everywhere at the same time. So in terms of maintaining brand consistency over a presentation, I think this provides a lot of opportunities.

I also just, I really like how the transitions work. I like that it has this Ken Burns effect for photos where it adds just this slight movement to the photo that if I'm giving a presentation, I feel a little bit of, an imperative as like the AI person that I'm supposed to have a slightly fancier presentation maybe than other folks.

This allows me to feel like I'm doing that without needing to, become an expert in, tools that are much more difficult to use.

And then I think this is a great sneaky little tool. Oh, I just had a question in on the last one. Can I track analytics on how people are using the presentations? Yes. you can. I don't use that functionality a lot, but every, you can make, like if you do want to track individual links, you can do that and you can see how many times.

That person with their email address has viewed the link. Or you can make, I had links made at IPW for when I was sharing different presentations at IPW. I can go to the IPW version of those links and see how many times those have been visited so I can follow over time. What does the access to the presentation look like?

And then there's a lot of stuff I haven't been using on sharing a password protective link. So whenever you share a presentation, you're actually just sharing a beautiful.ai like web address and then you view the presentation. I believe it uses JavaScript to create a flowing website experience.

So it's a lot easier, to send that than to send, a PDF of a PowerPoint and then trying to get that to be small enough to go over email. It saves a lot of that type of issues.

And we have another question on beautiful ai. Do they create landing pages? if you wanted to create one image and use that as a landing page, I don't think you could make it interactive though. But if you needed a single static landing page, I think beautiful AI could do that for you. I think they have a test period, but not a free tier, I believe.

what I like about napkin AI is that you can just paste in the text from a slide that you are working on, and then it will bring up all of these different ways that you can visualize it. And then once you select the way that you like. You are then able to come over and use all of these little tools to change the color, give it a transparent background to change the font here.

What I think is interesting is that these are the only words on the screen, right? But it's ai so it knows what I am talking about. So it move out of the way of this. So here is data security, protecting systems, data privacy, protecting people. I didn't tell it to write that. It knew that could add some useful context to the image it was creating for me.

So even if you don't end up using what napkin.ai gives you, I think it's really interesting to get ideas for visualizing information here. And then I will go back and I'll create it natively in beautiful AI if I don't like what Napkin created for me.

So then to sum everything up, if you were looking for one slide to take a screenshot of, just everything in one place.

I want to remind everyone that at work it is not cheating to use AI. Please keep using AI at work.

So now I'm going to go into the Q&A and start answering some questions. And we have a question from Tony about tools or recommendations for AI copywriting models and Tony.

I'm typically doing that in a Frontier model. So I will be exploring tools like ChatGPT or Claude to help me with writing a piece of copy. but I think Jasper is another tool that is very well regarded in that space.

Now we have a question about suggestions for a certification in AI for communications, something that would provide best practice with real world examples. I would say the AI Marketing Institute offers a variety of classes you can take and certifications. Allie k Miller also offers classes and certifications, as does Connor Grennan, and you can find all three of those folks or companies on LinkedIn.

Now we have a totally different question from Suzy, which is your information safe and secure on paid ChatGPT? What a paid version of ChatGPT gives you is SOC 2 compliance. It's protecting your data when it's in transit to the cloud and while it's at rest in the cloud. And then it gives you the ability to turn off model training. So for any of these tools, particularly if they're offering free access, means that we are the product. The way that we are the product for an AI company is that they need more examples of words used in sentences to train the next versions of ChatGPT, and Claude and Gemini.

And they can get that from us for free by taking all of the information we feed in and keeping it and using that to train the models. But honestly, we don't want our confidential information to be used to train a model. And so by using the paid models, you can toggle off the option for training on my data.

And that also helps to keep your data more safe and secure. But I would say the safety and security is not hIPAA level, so you would not want to put in PII, personally identifiable information like email addresses, into a tool. I would not put in information where in your contract it states you may not share the information with a third party because ChatGPT would be considered a third party in that scenario.

But don't use that to be scared off of working with these AI tools. Use that to reach out to the person on the other side of that contract and say, I'd really like to take your research and upload it into ChatGPT so it can help to guide my campaign strategy. Can we make an addendum to the contract that just outlines that under these circumstances I'm allowed to do that? Because the more of us that start asking for this, the more we get AI use out of the shadows.

And then we have some questions about agent ai. Keith asked, which AI does Agent.ai use? One of the coolest things about Agent.ai is that when you get to the steps, so the way that you build this, it's really building a workflow, right? One of the steps in your workflow for your AI agent is to take this information and put it into a language model with this prompt, and it has a little dropdown menu, and you can select pretty much any AI model out there.

Their vision for the platform is that if you wanted to build an AI agent and it was beyond your capabilities, you could find the person to help you at Agent.ai in their ecosystem. But also you could just use the agent that someone else made.

And every time you use someone else's agent, it uses credits and at a certain point you have to replenish those credits with your credit card. So that's ultimately how they monetize the machine.

But right now, if you are just playing with it as a builder, it's largely free to do that, you can use any AI model you want, which is really cool.

Once you start making those agents, does it allow those agents to be turned into swarms? The idea of an AI agent Swarm is that right now, maybe you can have one AI agent, or maybe you can string together a chain of AI agents. So AI agent that looks at information from a website and then maybe a second AI agent takes that information and transforms it into another use and hands it off to a third AI agent, which inserts it into Salesforce or Simpleview.

So in the future, some people are working on building a swarm system where one kind of supervisor agent could oversee hundreds or thousands of just single use agents that all do individual small tasks underneath the oversight of this oversight agent. So it is all a little beyond my capabilities at the moment, but I am learning about that stuff as fast as I can.

We're probably a little ways away from that being something that's easily applicable to our industry, but over the next couple of years, it's going to become not only more common to hear this talked about, but more common for regular people like us to be able to build this stuff because that's what's happening in AI is the need for technical expertise to achieve things, that need for that technical expertise is diminishing. And swarms are definitely one of those things.

We then have a question about an example of showing Agent.ai to enrich your email. The LinkedIn example would be a good way of doing that. If you just wanted to enrich an email with more information about who a prospect is, you can build an agent on Agent.ai that will pull from LinkedIn the exact information you would need to do that personalization and then write it for you to put into your email.

And there's other between n8N and Agent.ai, I don't think we're very far from, "okay, great. Now open up the email now send out the email on my behalf". But once we get into that piece, that raises a lot more security questions, then just having Agent.ai do something inside of its own ecosystem. So then that becomes, bring your IT team on board, bring your legal team on board.

I would have to do a great deal of research before allowing an LLM to have any access into our own email and calendar tools, right? Because that's where it can start to get a little dangerous when it comes to privacy.

We have a question about whether Lovable and agent AI can integrate, and I was listening to a podcast that talked about integrating all of these systems actually yesterday. So the idea that we could use Agent.ai to create backends, and then a tool like Lovable dev, which creates these really beautiful front ends and to combine those.

For the IPW project that I did a little bit of the coding, I couldn't get Lovable to just prompt its way to what I was looking for. So I actually mirrored the code base, in a code repo. ChatGPT told me what to do.

Say, all right, you're going to start by this and you're going to mirror it here, and this is how you make an account. You're going to look for this string of text and then you're going to type this over it in order to make the color different. The ability just to use plain language and not coding to do a lot of this work is very impressive and it's just getting easier.

Oh, we have a question about the Plaud note taker. Plaud is an actual little pin that you wear, or you can set it down, at a table during a meeting and it records absolutely everything you say and do during the meeting, which is great or unnerving. And I think for a lot of people that's unnerving. And I'll also warn you that during meetings, there are now a number of meeting note taking tools that do not visibly join a meeting.

So ChatGPT has a record feature that is not apparent to anyone else at the meeting. There's also a note taking tool called Granola that does not alert meeting attendees that it has been turned on. Communication's changing, and expectations around privacy in communication is changing.

So be cognizant of that as you use these tools, because I do think they're very helpful for meetings. I think it's very useful to be able to go back to a transcript and then to pull just the information you need out of that transcript to update a CRM. That's not even just the future, that's today, but we need to, take advantage of these tools in a way that's responsible and sensitive.

Thank you so much for everybody who, joined today and I'll get the rest of the questions sent to me so that we can address them in future webinars.

I really appreciate you joining. Thanks so much.