Generative AI features of Dynamic Chat

Hi, let’s look at the new Gen AI powered features that we are launching in Dynamic Chat. Let’s start with the experience that you can create for your web visitors. Here is a landing page. When a visitor lands on this page, Dynamic Chat automatically identifies if it is a known or anonymous and resolves their lead profile based on the cookie. A conversation is loaded based on that profile that you have targeted for that page. A visitor can start the conversation by clicking on the chat bar. As you might already know that you can enable visitors to go through a predefined flow, ask them any questions. Here it’s asking which Adobe solution are you interested in? Let’s go with Adobe Marketer Engage. Now with the Gen AI features, at any point in the conversation, you can choose to enable your visitors to ask any question about your products and services. Here I can ask anything about Marketer Engage. Let’s start with what are the key terms of Marketer. You can see it gave an instant response. This way, your visitors can ask questions about your products and services and get instant responses. Along with answer, it also provides related links that allow your visitors to get more information about the question that they are asking. You can also design a follow-up question that will nudge visitors to move forward with the conversation. A key part of this capability is, a chatbot will never answer any question or provide any response in a way that it is not intended or it is not approved by you. Let’s say here, a visitor asked a question that is nowhere related to your business or services that you provide. For example, will a visitor be back? You can see it says, apologies, but I couldn’t find information on this topic. Feel free to ask any other questions and follows up with a follow-up question. This way, you can enable your web visitors to ask questions about your business or services that you provide. Chatbot will always answer to the relevant questions with the responses that are safe and approved by your business experts. Now, let me show you how your live chat agent can also leverage these pre-approved responses during the conversation with the visitor. Here I’m back to the landing page where visitor was engaging with the chatbot. Let’s say visitor hits on this option, connect with an agent. It’s asking some qualification questions, so let me provide those. It’s asking for the e-mail address, nick123 at nima.com. It’s immediately resolving an agent for this visitor. As an agent, if once I accept this chat, I can see the whole conversation that has happened so far. To get the full context of this lead, I might have to go through this whole transcript to understand what has happened so far, so that I can avoid questions that have already been discussed. Or else, now with the power of Gen AI, it creates a nice small summary that has happened so far. Let’s see what it says here. The visitor is interested in Adobe Marketer Engage and inquired about its key features. The visitor’s name is Nick and their business email address is nick123. This is awesome. Not just that, it also identified the topics that visitor is interested in. It says customer engagement, Adobe Marketer Engage, and its key features. This is perfect. This way, LiveChat agents can get a quick context of conversation that has happened so far, while they are managing multiple conversations with different visitors. Let’s say agent here says, hey Nick, what would you like to know today? Let’s say in this case, visitor asks some technical question. For example, where does Marketer post data? If the LiveChat agent do not know the response to these questions, they might have to search for a resource or ask somebody to find the answer for this question. Or else, now with the power of Gen AI, he or she has an option to leverage Gen AI to generate response for them. So he can. They can click Assistant Response and it provides nicely the response for those questions that visitor asked. They can simply click on Copy and get all the response here. They can make any changes. Let’s say they don’t need the links to be added, remove them and hit Send. Just like that, now your LiveChat agents can respond to your visitor questions. Another interesting feature that we have introduced here is, let’s say the visitor has asked a complex question like this, which contains multiple questions in one response. An agent might know partial response to this question but not fully. So they can always ask the Gen AI by simply hitting slash. You can see these new shortcuts that we have added. Let’s say if you go by Ask AI, you can ask question to the Gen AI. Let me ask, what objects sync between Marketo and Salesforce? Immediately found a response from me, which I can simply copy and make any edits, which also includes the response for the other question, which is like, yes, make marketer is GDPR ready, and also the sync between, the rest of the response is beautifully crafted. I can further remove the links if I’d not need it, and hit Send. I can also leverage the other shortcut, which is Search. This shortcut will enable agents to find resources that they would like to share with the visitors during the conversation. Let’s say support links. Here you can see it provided me the links to Docs and to Dorys. Awesome. Let’s copy them, and agent can simply send these links to visitors during the conversation. This greatly simplifies agents handling multiple conversations, effectively and with confidence. Not just that, the summary of the conversation is generated for all conversations, be it automated chat or live chat. This conversation summary is pushed into Marketo through native activities, which means you can access the summary as part of the lead activity log. Here’s an activity log of a lead in Marketo. You can see the activity of dynamic chat which is engaged with an agent. If you click on this, you can see the conversation summary added to the activity along with other details. Also, we have introduced a new constraint called discuss topics. This allows you to target leads based on the topics they have discussed. For example, here, I can say discuss topic contains Marketo. This will give me all the leads who have engaged with any dialogue but discussed about Marketo Engage or Marketo. Let me show you how you can set up these Gen AI-powered experiences and other Gen AI-related functionalities. Once you log into dynamic chat, you can see we have introduced a new section called generative AI and go to Assisted Responses. As first step, we will leverage Gen AI to generate questions and responses from your business content. Let’s go ahead and click Generate Questions. You can give any task name. Today, I’m interested to generate questions and responses about interactive webinars and new capability that we launched in Marketo Engage. Let me give the task name as interactive webinars FAQs. Let me provide a link. Here I’m providing my documentation link. I can provide up to 50 links here. I also want to give the topic is about interactive webinars. I can add up to 10 topics. Let me hit Generate. What this does is, it takes the URLs that you have provided, extract the content from all those URLs, and generate the possible questions and responses about the topics that you have entered into the task. This task might take from a few minutes to anywhere up to 30 minutes depending on the number of URLs and the amount of content that each of these URLs contain. As you can see, the new task got created and it is in processing state. Once the process is completed, you will see a status called Completing, and also an option to download the questions and responses that are generated from the content that you have provided. If it fails, you’ll also see a message here, the reason why the task has failed. Either I can click here to download all the questions and responses in Excel, or I can go to Response Library. Here, I can see all the questions that are generated so far in all my tasks respective status and the task names and the topics that this question related to. I can click on any of the question, see what’s the question that got generated, and the response that got generated about the topic it is related to and the related read links. I can make edits to anything and set the status of this question as approved. Since I’ve already approved this, you’ll see the status here, and hit Save. Looks like a new task that we have created is already completed, and there are questions that are generated from this content. You can filter by topics, or you can even filter by the tasks, task names, and also the status of the question. This way, you can review your questions that are generated by generative AI from your content, and approve only the questions and responses that make sense to you. In case if you think you’re not the subject matter expert, and you would like to share these questions and responses to them offline and get them reviewed, you can always go back to the tasks and click Download. This will download all the generated questions and responses from the task into an Excel that you can share offline. Let me also show you the Excel. Here’s Excel, it has all the metadata, the task name, when it is got created, and you can see there’s a new tab called Q&Rs. If you go into this tab, you’ll be able to see from the link where the content resource and the topic on which the questions are created, you’ll see the generated question, generated response. You can mark them as the feedback here. You can select any of these statuses. If you think this is making sense, you can mark them as approved. If it doesn’t make sense, you can mark them as rejected. But if you think there’s a slight change that you need, you can simply mark them as approved changes, and copy the question, and paste it in the Edit Question column, make any changes as necessary, and that’s it. This way you can review any of your questions, mark them with respect to statuses, and save it, go back to your product. In the response library, you have an option to upload responses. You can simply click here, drag and drop the Excel file here, give it a task name, and hit save. It automatically updates all the questions and responses with respect to content and also respect the statuses. The good part is, all of this is logged and they are active. We’ll be able to see what are the tasks generated by whom and when. Next, let me show you how you can embed these generated responses into any dialogue that you design. Here is a dialogue that I’ve already created. You can see we have introduced a new card called generated response card. You can simply drag and drop just like any other card to the canvas. Give it a message, the message that you want to show to the visitor when they reach this point in the conversation. Let me say, feel free to ask any questions about our page. Here I can set the number of questions that visitor can ask in one go. Because our objective is not to just respond to visitor questions, but also move them down the conversation to achieve the KPI that you would like to achieve from these conversations. You can set the number of questions that a visitor asks in one go before they move to the next card in the conversation. Here I’ve set it as five, and then I can write my most question or the follow-up questions here, and provide my own responses. Based on the number of responses, you can add more options if needed, or you can remove them, you can also set the fallback message. This message will be shown whenever a visitor asks a question for which there is no response that is available in the response site. Once I’m done with this, I’ll go ahead and click Save. This is awesome. This created the generated response card with respective paths. I can simply connect the nodes to make the generated response card a part of the conversation. Here I have my first option which is connecting with an agent. Let me connect with this one so that it results to a live chat. If they have more questions, I will simply connect it back to the generated response card. If they are good now, then let me just connect it with this path. That’s it. All I need to do is just hit Publish. This dialog now allows visitors to ask questions and get instant responses when a visitor reaches this point. In Assisted Responses, we have another tab called Unanswered Questions. This will list all the questions that your visitors are asking to your chatbot, but chatbot did not provide any response. This will be a goldmine. Now you get to know visitors coming to your website, engaging with chatbot, and asking what questions for which you do not have any contact. You can select any question and simply click on it, write your own response, select any topic, add any related links of your name that you like, and hit Save. This automatically gets added to your response library, and next time when visitors are asking about this question, they will be provided with a response. These are all the features that are powered by Edge and AI that we launched as part of the phase 1 rollout. We have many more exciting features that are coming up into Dynamic Chat. Looking forward to see you leveraging Dynamic Chat for your business.