AI Assistant in AEM

Transcript

All right. Welcome everyone. Let’s give some few minutes to people to start rolling in.

Great. Give it another minute.

All right. Good morning, everyone. Good afternoon. Good evening, wherever you’re at in the globe. We have a great session prepared for today. We are going to talk about AI Assistant within AM. So we have lots of great content for you. This is one of those features that we are so happy with an Adobe to share with you. And let me kind of walk you through who’s going to be delivering this. So, Arshad Das is our principal field engineer within field engineering. He will be walking us through this feature. He’s going to walk us through the content. I’m going to actually be co-hosting, basically trying to answer any of the questions you may have during the chat, as well as I’m going to be moderating some of the webinar as well. So I am a senior field engineer within field engineering as well. We both have AM backgrounds as well as just to get for performance marketing. And we’re going to be diving into what we’re going to be seeing today. So, Barshan is going to tell us exactly what AI Assistant is. All the supported solutions within the experience cloud is going to tell us all the prerequisites for the solution. He’s also going to talk to us about configuration and setup, best practices. He’s also going to walk us through a demo. We’re going to be as well looking at some of the references. So he’s going to provide some of the references as well. We also have a couple of polls that we are like for you to participate on. And actually, let’s go ahead and start with the first poll before we start diving into the content. So this first poll that I would like to present is just your general adoption. Are you familiar with AI Assistant at all? Have you played with it? Do you know what it is? Please let us know. We are very happy to kind of get your intake and learn more from you. So please go ahead and submit it.

Let me know if for some reason the poll did not show up.

All right.

It looks like most of you have heard it, but never used it. So I think this is going to be a very useful session for all of those that have heard about the actual offering, what it is and how you can leverage it. So you’re in the right spot. This is exactly why we’re having this webinar. So you can learn more about these offerings, learn more about the usage and how you could leverage them. So with that, I’m going to pass it on to Varshan now.

Yeah, thank you, Marco. Hi, everyone. So as Marco rightly mentioned, in this session, we are just going to talk about the AI Assistant in AEM. So, as you know, with all the AI tools that are coming up, most of the AI Assistants are getting embedded in each of the products in the Adobe Ecosystems. And today we are just talking about the AI Assistant in AEM only. So this AI Assistant provides you a conversional companion embedded within the AEM. It is built on the AEP Agent Orchestrator. So if you look into the AI Assistant in details, and I’m going to show you as part of the demo, you’ll see that it has got two purpose-built agents. One is for the product knowledge and another is for the product support agent. So with the product support agent, you can actually interact like a normal conversation where you tell the agent what to do with a particular support ticket. You can look into the status of the support tickets. You can also create a support ticket within the AI Assistant that we have. And of course, the product knowledge where it tells you details about a particular product. So there are certain products that are supported and we can go into that one later on. But these are the main two built-in agents that we have within the AI Assistant to cater to the product knowledge and the product support agent. It does provide you an intent-aware help. So it tries to determine what is the intent of the conversation and provide you a response accordingly. If you have asked a question which is not really within the scope or within the boundaries of the AI Assistant, it will also tell you that. And as most of you have already known by now, it is embedded within the AI Assistant in itself. And there are different peripheral products of AI where you can actually get to see this particular AI Assistant. Now, what are the key benefits that you get with the AI Assistant? So just like any other AI Assistant, one of the benefits is to speed up the time to value. So if you look into the timeline from the time to value perspective, it is eventually going to speed it up because you are now going to get the immediate answers to a particular product knowledge question that you have immediately within the AI Assistant. So from that perspective, it speeds up the time to value. It also cuts down the support load. For example, imagine you are already within the UI of the AI model in any of the peripheral products. Instead of you going to another screen or opening up the support portal, you can right there open up a support ticket. You can also know details about the support ticket. So all those things are going to help you in cutting down the load from the support or the time that you spend separately. So from being efficient, that is one of the key benefits as well. And it makes AI more intuitive for you. So while you are navigating across certain features of an AEM as an author, if you want to know about a particular feature, you can immediately open up the AI Assistant and ask it and it will tell you the details about that particular feature. So you did not scroll through the Experience League articles to know about it. Of course, it actually crawls the Experience League articles and it indexes it at its own level. And that’s how it is able to show you all those responses. But it saves a lot of time for you to start actually digging out for a particular feature, know how it works, to directly ask the AI Assistant how it’s going to work and it will tell you all those details. So based on the intent awareness it has built in. So these are some of the key benefits that you are going to get using the AI Assistant in AEM. Now, let’s look into some of the supported products. So Marco, if you can go to the next slide. So these are some of the supported solutions for the AEM AI Assistant. So when I mean supported solutions, one is the interface from where this particular AI Assistant can be invoked. Another thing is what are the knowledge bases that it’s going to support. So when we are talking about the AEM AI Assistant, these are the knowledge bases that it’s going to support. Whether you ask the Assistant questions about the experience of the delivery services, AEM size, assets, forms, dynamic media, cloud manager, these are the knowledge bases that it can easily support. Now, if you ask a question which is outside its scope, for example, if you ask a question about a Figma plugin or a Photoshop plugin, you will see it in the responses that it will tell you very specifically that it is outside the scope of the AI Assistant. So in terms of supported solutions, these are the supported solutions. As of now, we can also tell you that there are a lot of innovations happening at this end. So a lot of things will get built in over here over a period of time. But for now, these are the supported solutions that we have for the AEM AI Assistant.

Now, let’s look into some of the prerequisites that are needed to set up the AEM AI Assistant. So one of the prerequisites is the customer should be signing the Gen AI Rider that will give you access to that AI features.

Once this rider has been signed, the AI Assistant feature is turned on by default. And once it is turned on by default, it is just about ensuring that your user ID or the user email is actually part of a relevant group and eventually part of a particular product profile. So that you can access the AI Assistant.

And if you have to set it up, you need to have the product administrator rights at a minimum level from the Adobe Admin Console. And I’m going to also walk you through the Adobe Admin Console to show you how you are going to set it up. It’s a very streamlined process, just a few steps earlier. But yes, you do need to have the product administrator rights to set it up or to create the groups because you have to enable other users if you have to enable other users.

And one requirement, it’s not a technical requirement, I would say is understanding your organization’s user management structure. Right. So you first have to decide as in who all are the users whom you want to enable the AI Assistant for. Preferably, you would definitely would like to enable for all the AEM authors as an example. Right. You would also like to enable for the administrators or maybe if someone is into the cloud manager, someone having the deployment manager profiles. So you would definitely solve those things are needed. But yes, it’s a decision to be taken at each organization’s level that who do you want to enable the AI Assistant for what all users. And based on that, you will be adding that particular user or the user email into the relevant group, which would eventually be associated with the product profile for the AEM AI Assistant.

Now let’s look into briefly into high level, what are the different configuration and the setup.

So at a very, very high level, these are the things that you would end up doing. Right. So you have to create a new product profile. You can also make use of an existing product profile, but it is generally not recommended because you want to have it separate. So ideally, it is recommended to have a new product profile in the admin console. And then you enable the AI Assistant product knowledge permission within that particular profile. You can also, you’ll also have to create a new user group. So that helps you identify very clearly that this is the specific group for the AEM AI Assistant. And once you get into the group, you also get to see all the users who have access to this particular assistant. And then eventually you’ll be adding those users into the user group and assigning a product profile to the user group. So that is how it is eventually going to work out for you at the high level.

Now, before getting into the demo part of it and showing you how it actually works, it’s also important to understand certain best practices that we recommend while you are interacting with the AEM AI Assistant. So just with any other AI tools, it is very important to provide clear inputs, right? So the response of the assistant is actually going to depend on the inputs that you provide. So if you provide very clear, specific input, it’s going to be easier for the AI Assistant to understand and it will respond accordingly. You do have an option to refine your prompts as needed. For example, you asked for a particular topic or a particular product knowledge. You think that based on the response you want to drill down again further, you can refine that prompt as needed to get very specific response from the AI Assistant. Now, when I’ll be showing you the demo, I’ll actually be walking through some of the false positives as well as in some of the cases where the response won’t be there or it is not idle. And as you know, and I talked about it before in the previous slides, the AI Assistant consists of two built-in agents, right? One is the support agent, another is the product knowledge agent. Now, other than that, if you are trying to look into very specific, for example, if you’re into the AM cloud manager, a cloud manager pipeline related questions, it won’t be able to answer. Right? So when it comes to the product knowledge, it can answer all the product knowledge of the supported solutions and the products that it supports and anything related to the support ticket that also it will be able to answer. But very specifically, if you want to ask about a particular pipeline, it is currently outside the scope of the AM AI Assistant to talk about those specific details of a particular pipeline. Like, for example, if you want to ask why this pipeline has failed, so that currently is out of scope. So the best practice is always to know about what all things the AI Assistant is currently supporting and refine your prompts accordingly. With that, we must also mention that it is always recommended to evaluate the output. So when you see the response from the AI Assistant, you also get to see, most often you also get to see the sources that it has mentioned. So do look into the sources. Looking into the sources will give you a better confidence level on the response. Most often these sources will be from the Experience League articles. But it is also recommended that you evaluate the outputs that are being provided from the AM AI Assistant.

Okay, so at this point of time, I think we got a quick knowledge of the AM AI Assistant and I’m going to move directly into the demo part of it. Should be a quick demo. Let me share my screen.

Okay.

I hope everyone is able to see my screen.

Yeah, so this is the admin console of one of my sandboxes that I have. And from here, I’m going to first show you how you can set up the AI Assistant. But before even before getting into that, let me show you this is where the AI Assistant is. Again, most of you have must have already seen it. So this is basically the system that we are talking about. Right. So if you go on the right hand side, you click on this particular icon, you get to see the system coming up over here. Now, how do you set it up? Over here, if you see in my admin console, I do have the product administrator rights and I have created a particular user group.

Let me directly take the name of the user group and put it up over here.

I’ll go under the user groups and just search for the particular user group that I have created.

So this is the user group that has been created. Currently there are eight users. Now when I look into it, so all these users that I have added as part of these user groups will have access to the AI Assistant that is in there. Right now under this user group, if you look into the assigned product profile, it will point to a particular product profile in the cloud manager.

So this is the product profile that I also created. Now let me go under this particular product profile.

So you see this is the product profile under the ABM cloud service cloud manager that I created. And this product profile is directly associated with the user group that you have. The key thing to see over here is the permissions. So just like any other product profile, it will have a lot of check boxes to let you decide what all permissions you want to give to this particular profile. Now the one that is important is, let me go to the permissions tab, is the AI Assistant product knowledge. So this has been turned on.

So this actually has been turned on and this is what basically enables the AI Assistant for the relevant users for this particular product profile. And whoever users are eventually using this product profile or are part of this product profile will have access to the AI Assistant responses.

So over here you see that it has been enabled. I had also enabled other things as part of this one, but this is not really needed or recommended. If this product profile is actually only for the AI Assistant, you might as well enable only this part.

So this is the basic setup that is required for you to enable the AI Assistant. And of course, from the customer’s perspective, you have to sign the rider agreement. Now let me take you to the different prompts and show you the AI Assistant in action.

So this is where you click to get the AI Assistant and I have already saved certain prompts. That is going to be useful. Just give me one more.

So there are some general product knowledges that you can make use of. For example, if I want to ask about a particular product like the Content Hub, I put in the prompt over here, like how to manage access for the Content Hub. It will be able to show these responses over here. So it gives me all these details, like how I can manage access to the Content Hub. So even though I am within the cloud manager interface, because this product profile also has the built-in agent for the product knowledge, and this AI Assistant actually supports four or five different solutions within the AM and the peripheral products of AM, it is able to answer me those questions about Content Hub as well.

Similarly, for the forms, if I want to check.

You can see that it is able to answer the details about a particular forms related questions as well. And if you look into the sources, it will give you the sources. Most of the time, these sources will actually be from the Experience League article, which tells you from which Experience League article it has actually passed those responses from.

Now let me try to show you, for example, if I want to understand the difference between dynamic media, right? So we tried one for the Content Hub, we tried one for AM forms. Now let me try to show you about the dynamic media scene 7 stack and open API stack.

So this is the response it is going to show you about the scene 7 stack and the open API stack. So you will see that these are all the solutions that it supports and different kinds of product knowledge questions that it is able to answer. Now you can also get into a little bit more details.

For example, if you want to know about what are the implications of excessive content nesting in AM, let’s see how it is going to respond. So it is more on the technical side and getting a little bit more into details than just generic product knowledge. And you can see even when you specify a detailed question, a detailed technical question, for example, implications of having excessive content nesting in AM or the hierarchical structure in AM, it gives you the response pretty correctly. So one of the things is about the performance concerns, increased query complexity, content governance challenges and authoring experiences, or the complexity around authoring experiences. So these responses are pretty much correct. And you can see it’s not just about a generic product feature based knowledge, but also a little bit more technical knowledge is within the AM or other products that it will easily be able to answer.

Now let’s ask a question about the licenses. Just give me one moment.

For example, I want to ask whether an AMSS license can give me access to the dynamic media with OpenAPI Lite and let’s see what it responds to.

So you can see it is also able to answer some generic questions with respect to the licenses. It is not specific to your license that you would have. For example, if I ask how many licenses do I have or how many power users I have for this particular org, it won’t be able to answer that one. But whenever it comes to generic questions with respect to whether you need to operate dynamic media with OpenAPI, it will just say AMSS license would suffice. It can answer it accordingly.

Similarly, for edge delivery services, for example, if I put up what are the different benefits of using edge delivery services, it lays down all these answers and these are pretty much correct. So these are some of the examples of supported solutions that I wanted to show that the AI assistant can actually answer pretty much correctly for you. But if you are trying to get into very specific details, for example, if you want to say how many customers are currently using the edge delivery services, it won’t be able to answer because this is not part of a generic product knowledge. And you see how the response would come up that currently the AI assistant’s knowledge is limited to AM concepts.

So this is how it is going to come up. Another example that I wanted to show that is going to be outside the scope of the AM assistant, for example, a different knowledge about a particular product that is not supported by the AI assistant. So this would be it. So, for example, if I want to know about what is the latest plugin in Photoshop, let’s see what it responds to.

So it clearly tells you that it does not provide specific information regarding the latest plugin in Photoshop. But at the same time, it also suggests some other questions that you may ask. Right. For example, if you want to ask, how can I install a plugin in Photoshop? Let’s see what it says. It says it needs additional verification. Right. So most often, if you see if you’re trying to ask a question which is outside the scope of this particular AI assistant, it will give you a response that it is outside the scope and it will also give you a warning message that it needs additional verification.

Now let me get into the AM assets plugin.

So over here, you see that I have asked a question about the asset plugin of AM assets, which is one of the supported solutions and it is able to answer it correctly along with the sources that it has identified.

Right. So as long as it is part of the supported solutions, it would be pretty much be able to answer any of the generic product related questions that are in there. So this was about from the product knowledge perspective. And I can also show you. So just as you have it in the cloud manager interface, you also have it in the AM interface, right? It’s basically the same AI assistant that is in there, but just a new instance of it. Now, if you want to start this AI assistance from the fresh, you can do a start new conversation.

And it will consider this as a new conversation for you. Right. So over here, let’s say I want to know. So I had created a support ticket from this AI assistant itself. And I wanted now it has been closed, but I wanted to actually ask about the status of the support ticket. Right.

So this is the support ticket. So you can see it clearly tells me the status of the support ticket.

Now, let’s see if I want to be more specific.

Let’s see what is going to respond in this case.

Yeah, so you see that if I ask the reason why was the support ticket closed, so there could be multiple reasons associated with it, right? So what it responds back with is the title of the support ticket and the current status associated with it. Right. You do have an option to also create a support ticket from here. And let me try that one. So that response takes a little bit time, but I’ll try it from here.

Let’s say I just try to create a generic support ticket stating create a support ticket for my cloud manager pipeline getting stuck.

It will tell you that you can review and edit the ticket. Whenever you click on it, it opens up an interface for you. Right. So over here, you actually determine that whether all these details are correct. Right. So it populates certain fields for you. For example, the product and the product offering the environment type. Let’s say select the environment type is development.

Right. And based on this, I actually generate the support ticket. So I do have an option to actually change the description to put up more details. One thing to note over here is at the end of the description that is going to go into the support ticket, you have an AI assistant chat ID. Right. So your support engineers, whenever they are looking into the support ticket, can easily identify how this particular support ticket has been created, because it has a chat ID associated with it. Right. So with this one, I’ll go ahead and create the support ticket.

Okay, so by now the support ticket has been created. So this is the support ticket. You can see the status. It says awaiting assignment. Now let’s try to immediately pull in the status and see if it reflects the correct status or not.

So I just mentioned what’s the current status of the support ticket.

And it gives you the exact status. Right. So this is what I wanted to show you. You do have the ability to directly create the support ticket from here. And it’s not that whenever you are entering, it will immediately create it. It will let you to review and edit. So if you specify more details, those details will be pre-populated while creating the support ticket. Otherwise, you anyways have that option to go under the review screen that I showed where you’ll be able to select the relevant product. If you want to change the description, you can change those descriptions as well and put in more details and then create the support ticket.

So yeah, so over here it is also giving me the status of the support ticket. I’m happy with the response. So I’ll just go ahead and say thank you.

So these are the two main use cases that I wanted to show about the AI assistant. The first one that you actually saw was about the product knowledge, like how you can actually prompt the AI assistant to know about different product knowledge and the features and details. How you can also go into little bit, even though those are generic questions, more from in terms of technical terms in AEM. How can you actually ask some of the technical questions to the AI assistant and what are the things to avoid that will be outside the scope of the AI assistant. And then finally, this use case where you can actually interact with the built-in support agent that is there to create the support ticket, to add the status of the support ticket and basically maintain the support ticket, without getting into separate screen or the support ticket portal.

So yeah, that’s pretty much the demo that we have. And at this point of time, I think we can take some questions. As part of the document, I have also put up some of the reference link, which will tell you all the steps that I have shown. So you will have step-by-step guidance as in how you can set it up. The one that I have shown over here. What are the best practices you would also get to know over here. So the general overview of the AEM AI assistant is there. You also have the configuration steps. And it is also important to understand what are the generic user guidelines. When it comes to the Adobe Experience Cloud generative AI. So I have put up a link for the AI user guidelines as well. So do take a look. This deck will also be shared with you. But I think at this point of time, we’ll take a pause and take any questions.

Before we dive into the questions, by the way, thank you, Varshan. That was very insightful and very powerful tool. Thank you. It was great to actually see it in action. Before we actually dive into some Q&A, I would like to present a poll real quick.

And while you start kind of thinking about some of those questions.

This one is around, you know, the understanding of the solution that was actually presented to you.

As well as we’d like to know a little bit more about the things that you would like to see in these webinars. Please do share your thoughts and ideas.

We’re welcoming, you know, topics pretty much every month. So please do share with us.

Look and see.

See the results.

Are you trying to see how to view the poll results with this? All right.

Okay, with that, let’s go ahead and dive into your questions. I know we answer a few already on the Q&A.

If you do have a question, please post it on the Q&A tab.

Yeah, and I also just want to answer this particular question. I know that Marpih have also answered. So this is from Julie who says, in my company, not all users can create support tickets. Will the permissions be inherited in the AI assistant and make sure that no generic end user can see? Yeah, so there are basically, in addition to what Marco said, I must mention that there are basically two types of permissions that we are talking about, right? One is permission to the AI assistant that can be controlled, what I shown. Another is the permission to create a support ticket. So when it comes to the permission of the AI assistant, whether someone will be able to prompt an AI assistant, it can be controlled using the product profiles that I shown as part of the demo. Now, when it comes to the permissions of creating a support ticket, that is determined based on if that particular user ID or that user is part of the support group or the support product profile. So for example, if I’m trying to put up a support ticket and I do not have access to create a support ticket, the AI assistant will stop me. So even though I’ll have that ability to ask generic questions, if I want to know the status of a particular support ticket and I do not have access to create support ticket, then it will be stopped. You’ll be able to see it in the response in the AI assistant.

Great.

I don’t see any other questions.

I did not see any other questions from attendees.

And I do see that the polls are, the participation of polls is great. It’s amazing.

Yeah. And we would also like to highlight one thing, like in future, a lot of new features may be added over here. But for now, as I talked about repeatedly on this one and always stress from this particular point that it is about the product knowledge and the support knowledge that the AI assistant is currently catering to. In future, a lot of other things and innovations may happen and will be brought up as part of the AM AI assistant. So that is not something that we can rule out as of now. So if any one of you have questions and then what all things in future will this particular AM AI assistant will be able to support. Yes, it is still in the works. We haven’t got a concrete roadmap yet.

And so, but currently it is about the product knowledge and the support knowledge for the support tickets. And yes, your interactions are based, for example, your support ticket interactions are definitely based on the access that you have. For example, if I don’t have access to create a support ticket, the agent, the AI assistant will stop in between and say that I don’t have that access.

We do have another question from Veronica. Are there any uses for content generation? Yes, that’s a very good question.

Yes, that’s a very good question. So it is not there yet. As I said, it is basically your product knowledge and the support related stuffs. But that doesn’t mean that it won’t be there in the future. I know that a lot of things are in works, so which we cannot explicitly mention over here on this call. It’s better to hear it from the product managers in particular and you will hear them when it comes. But yeah, there are a lot of things that are in works. But as of now, it is only the product knowledge and the support knowledge.

We also have another question from Bury. Does it support content authoring in any way? It does not allow you to directly author content from here, but it supports any knowledge that you want to know about content authoring or content generation.

So if you look into, for example, if you divide what all things you can do with this AMS system to two different things. One is read access, another is write access. So in terms of write access, it’s basically the creation of the tickets that you have. All the product knowledge that you get are part of the read access. That doesn’t necessarily mean you will have read access to everything. For example, if I say for this particular AMS org on this particular pipeline, can you read this log? It won’t be able to do that, right? Even within read access, there are certain limitations that it has, which I have shown as part of this demo. But yeah, just to answer your question in short, no, it cannot directly author or generate content as of now.

But yeah, currently it is based on those two built-in agents that I talked about.

That’s a very good point because those agents don’t have access, direct access to any of your environments or repositories. So if you have any specific questions about your code base, it will not be able to dive into that. Let’s group on.

All right. Anyone, any more questions? Okay. Well, it doesn’t look like we have any more questions than we could definitely end earlier.

So feel free to go and go through the documents or the resources that Barshan mentioned. There’s a lot of documentation on Experience League about AI Assistant, how you can use it, how you enable it, everything that has been presented here. You can find it also on Experience League.

Feel free to work if you are already working with a TAM, be working with anyone within Adobe. If you have further questions, please feel free to add those questions to them or they’ll be able to assist you. All right. Since we don’t have any more questions, then I think we are done.

Well, thank you very much, everyone. Thank you, everyone. Thank you.