Name: Scale Up: Content Generation in Yext
Uploaded: 2024-09-26T16:52:15.828Z
Duration: 20 min 4 s
Description: Scale Up: Content Generation in Yext

Transcript for "Scale Up: Content Generation in Yext": Hi, everyone. Thank you for joining the session Scale Up Content Generation in Yext. Today, I'm here with Trent, who is our product marketing manager, and I'm product manager here at Yext. And we will be going through content generation in content as well as reviews to give you guys insights as to how Yext can help you scale. Today, customers expect content in more places than ever. You might have a blog post that is fit for your website, but you also need a summarized version for an email blast, or you want to have, FAQs for a chatbot, and you want to verticalize that content for your different types of customers that might range in ages or they might be for different industries. And you might have the basis of that content, but how can you efficiently and quickly use that foundation to create more and more mediums to share that and make sure that you have content that is best fit for all audiences? Generative AI is great because it can help unlock that potential to scale efficiently. It can help you save time, and you can then prioritize more important work. However, generative AI in general has a lot of risks. For example, generative AI is not always accurate. It's known to hallucinate if you don't provide it with the foundational information that is going to make sure that there is no hallucinations that it can make up information about. For example, if I want a blog post about my business, I need to make sure that AI knows a little bit about my business before it goes off and creates that content. In addition to accuracy, quality control is definitely important. Humans should always be approving content before it's published, staying in the loop. And if we're gonna have humans involved anyway, we want to make sure that we can still have that efficiency aspect. So how can we save time, use generative AI, but also ensure that our content is accurate and of high quality? That is where Yext can help. Yext can help mitigate risks in 3 simple steps. Number 1, Yext allows you to store content in the knowledge graph, and that content can help you generate content as the foundation of the prompts that you use when you're generating content with AI. This creates accurate content. So you already have all of your fields filled out, for example, in an entity. Like, you have your business hours, you have your, description, your address, your keywords, and how can you then use all of that information to create a business description or a blog post or something that is using that foundational data that you already have recorded in Yext. Secondly, efficiency. If you have all that information and you need to just generate one business description, you could use chat g p t and copy and paste that information into a prompt in Yext, but that's not super efficient. Yext can help because you can write one single prompt that can apply to 100 and thousands of entities at once. I'm gonna show more about that in a demo, but that is why Yext can really help drive efficiency in using AI to generate content in a way that is dynamically using information that is unique per entity or per location, but you can do it in a single click. Lastly, Yext can help because we have suggest a suggestions framework that helps ensure quality control. Any content that's generated, if it's automatic or if it's doing so in bulk, can go through a suggestion framework so that way a human can approve it, make changes, and then publish it when they're ready. So summarizing that process altogether, users in Yext can configure a prompt in order to generate content. That content can be generated at scale and use suggestions. Humans can then approve, modify, edit, and then, finally, they can publish that content in however many mediums they want. So that way, it is accessible in every form that you could possibly imagine. Hey, everyone. Thanks for tuning in to the demo portion. I have navigated to the knowledge graph or content in Yext, and I am on an entity that is a help article titled The Role of AI in Improving Business Operations. Fun fact, this help article body content was actually written by Chad GPT. And what I will be demoing is how we can use that, body content that I have written. And in a in a scenario, like, for a real business, maybe you have written help articles for how to navigate your own systems or help articles for employees to use in Internet search experiences. And now that you already have that foundation of knowledge, how can we scale that knowledge to then create content for different mediums? So you might want summaries as mentioned or verticalized information, keywords. And so I'm gonna go through how you can use Yext to do that really efficiently. So this body content again has already been copy and pasted and lives in Yext as a field. And I'm going to now click this wand, and this wand is going to generate keywords off of this body. So these keywords can be helpful maybe in, SEO optimization on your website. And what I can do is I can make any changes I want. So maybe I don't like this keyword and maybe I wanna add another one that is, like, AI concerns, let's say. And I can then click save. Another example is generating a summary or a short description. Maybe a summary is really useful for email content or because you have an audience that just doesn't have a lot of time to read a full blog and you wanna still appeal to that, to that audience. I can then click the one. I can click it again. It's gonna generate content again, and we can click save, edit, etcetera. As mentioned, another use case can be generating meta descriptions. So this can be used in the HTML if you're putting this content on a webpage. And some other examples is actually transforming this entire body content into a verticalized version. So I would like to have a health care version that is specific to my health care customers or a financial services version. So that way, you make sure that every single audience is feels targeted and feels important. But, again, all I needed to do was have this one foundational piece of data to then create the health care, version of that. And this is still generating. Oh, and there we go. So you can now see how it tailored it directly to healthcare. So we have, some key inform some keywords like healthcare providers. It's gonna make sure that it talks about the healthcare use cases, AI managing patient inquiries around the clock. It really focuses the importance of AI specifically on the healthcare use case. So the aspect of using the content that you already have in the knowledge graph to make sure that you are creating accurate information can be seen here because the AI is not just going to hallucinate creating keywords about the role of AI, but it's actually gonna be able to take keywords using the data you already have. Another example could be creating FAQs from this body. Instead of having AI just say write me FAQs, you already have so much data here. So maybe you also have FAQs about your business hours, and you want to use that that foundational knowledge to then generate the FAQs. That can all be done right here. And how do we do that? In the fields, you are creating these prompts, that are then saved to the field configuration and applicable whenever you are, generating contact for that field. So the prompts that I have for the health care example is rewrite the following article for a target audience of the health care professional industry. You can see here that we've embedded the body field. By embedding other fields on the entity, that's how the AI is able to look to all the fields or any field that you choose and be able to, create new content based off of what you already have. So another example could be, keywords. Here, you can see we have a list of, also built in prompts. So if you don't know how to write prompts, that's okay. You can absolutely just use a pre built one for some of the common use cases that we've seen. And for keywords, you can add a topic, you can add a desired number of keywords, and some additional instructions. So maybe you wanna say, look at the summarized version of this and pick out some summer some keywords of the summary. Finally, another, great aspect of ContentGen and Yext is that we can make sure that your content in all your mediums are kept up to date. If I make a change to my original body content, I wanna make sure that the verticalized version or my keywords are also changed simultaneously. By having automated computations, any of the embedded fields that change can also induce, cause the field content in the generated field to regenerate content. So in the example of the, verticalized information, if I make an edit to my body content and I just want it to be, like, leveraging technology for better I just wanna make it shorter and I wanna take out this piece because I don't think it's really important. I can then save this and editing this body will automatically or it can automatically if you choose to toggle that feature on. It can automatically regenerate content in those other fields. That content then applies, either directly to the profile or tying back into what, Yext offers as the quality assurance piece is it can apply as suggestions. So you can keep that human in the loop and make sure that somebody is ensuring quality. So we've now really covered the quality, the accuracy, and finally, getting into scale. Not only, can you generate content by just clicking that wand once on that single entity. As a reminder, that's how it's done. But you could effectively click the wand on tons of entities and in a single click. I can do that by navigating to the content screen. I can click trigger field computation, and let's say I wanna trigger it on the keywords. And I have all these settings, so I could do it on all of my entities. Let's say I want to apply it as a suggestion. And then when I click compute, that's going to trigger that keywords computation on all of my entities, and it's going to make sure that the prompt is dynamically applied to every single entity. So that way, the keywords of help article a are relevant to that entity, and the keywords are generated uniquely for help article b. Again, suggestions tying that back into the quality assurance. We have accuracy and we have scale, and that's where Yext really stands to help you, use generative AI to scale with all of the safeguards in place. Thanks so much, and I'm now going to kick it over to Trent to talk about generative AI in review response. Awesome. Thanks so much, Rachel. I'm gonna talk about another use case where you can use generative content to improve your productivity, and that comes in the form of responding to customer reviews. And before demonstrating how generative review responses can enhance your efficiency, it's essential to first understand the importance of responding to reviews. We can see here reviews are a critical part of the customer journey. Almost every single person reads reviews before making a purchase. And on top of that, 94% of customers say that reviews have made them avoid a business. To improve your online reputation and that brand trust, one tactic that you can use is to respond to your reviews. And the data shows that the more reviews you respond to, the better. Brands that respond to at least 75% of their reviews are rated more than half a star higher than businesses that don't respond at all. Half a star increase can not only impact perception of your business from people who are searching and engaging with your brand, but it can also have a huge impact on where you show up in search. And that's because search engines rely so heavily on ratings and reviews to help determine how to rank businesses in search. Sites like Google will look at a number of factors, including the quantity of your reviews, your star rating, how recent your reviews are, your response rate, as well as the diversity of your review sources. With reviews having such a big impact on how customers both perceive your brand and where you show up in search, managing and responding to customer reviews presents a huge opportunity to boost your brand's reputation and drive additional revenue through enhanced search visibility. However, the challenge that most brands face is dealing with the amount of time and effort that is required to effectively respond to reviews. It could take time to craft a thoughtful, empathetic reply to an unhappy customer. And when you multiply that by 100 or even thousands of reviews, depending on how many locations you might have, responding to reviews can seem like a daunting task. But now you can use AI to craft authentic review responses in a fraction of the time that it would take to respond to a review manually. Not to mention, you can add instructions to the prompt to make sure that responses are always in your brand voice. So now let's go ahead and show you exactly what that looks like in action. Alright. Here we are. I'm now in the X Platform under our reviews product tab. Here, we can see all of the reviews that are being managed across the web. Those are being pulled in, and here, we can manage and respond to those in real time directly from one place. At the top, here are some high level metrics to call out. So these are the number of reviews being pulled in, the average rating, breakdown of sites, as well as the number of reviews, for each star rating. I can quickly click on any of these to filter. So any of these metrics at the top here, or I can use this filter bar at the very top to search for any type of criteria. So if I only want to look at all of my Google reviews that are, let's just say, less than or equal to 3 stars because those are the reviews that I care about, we can apply this filter. And here we go. We have all of our Google reviews that fit that qualification. I'm gonna go ahead go ahead and reset this here. But to go ahead and to respond to review, you would just click on this response button over here on the bottom right hand corner. Now, of course, you have the option to manually respond to this review. There's not a lot of context here. Unfortunately, this was a negative review. But instead of manually responding to this, you now have the option to click this one button that says generate, and that will craft a personalized response to this review using AI. So you can see here we have a much more thoughtful, empathetic response. Right now, this is the prompt that is being used. So if I actually go ahead here and click view prompt, this is the instructions that I have provided this model to respond with a tone of sincere, caring neighbor, and add an emoji at the end there to to add a little bit of fun. But, you can provide any sort of brand guidelines or instructions that you wanted to follow. So if we wanted to edit that in settings, this would take us directly to where we can now provide those further instructions and those guidelines. And that way you're able to provide these responses with your tone and, of course, using your brand voice. So if I wanted to remove this part here at the remote the emoji, I could then just hit save, and there we go. I've quickly updated our, our model to now, provide more thoughtful instructions. So we're gonna go back here over to our reviews that we were just responding to. Right now, this is the prompt that I have instructed the AI model to use when responding to reviews. Here we have respond with a tone of a sincere, caring neighbor, add an emoji. Let's just say we would like to remove the emoji from the instructions here, but you can provide any sort of guidelines or instructions that you wanted to follow. So that way your responses are always crafted using your tone and brand voice. And not to mention, you can free text these responses as well. So if I actually go back, and here we are back in this review, go ahead and generate another response. It will update based on the revised instructions that we've just provided. So now if you wanna make tweaks or changes, you can go ahead and free form text. So let's just say we wanted to add or update anything, but in this case, I like exactly how this response is. To officially respond to this review, you would just hit this respond button here, and that would directly post to the site in real time. Right now, because I have full editing access in this account, I see this respond button. But let's just say you wanted to have someone else approve this response before it got published. You could set the correct per permissions so that way the person creating this response with AI could see a button that says submit instead of respond. And, therefore, once you hit that submit button, that response would instead go into an approvals workflow or anyone else on your team could approve this response before it actually gets published. So whether your team decides to set up rules or any sort of approval workflows, using AI to respond to reviews is a game changer because it allows businesses to create more personalized and thoughtful responses at scale. So now you and your team can reap all those benefits of an effective review response strategy without having to expend a lot of time and effort, especially for those of your marketing teams that are very lean and already responsible for so much as it is. Well, thank you so much for joining Rachel and I for today's session. We hope you enjoy the rest of Yext Summit.