Next-Gen customer experiences & efficiency with Adobe Experience Platform Agent Orchestrator

Transcript

Hi, I’m Daniel Wright, Senior Technical Marketing Engineer live at the Adobe New York Union Square office. I’m excited to host this episode of Experience League Live, Next Gen Customer Experiences Platform Agents T-Shirt. That’s how you know we’re serious. Whenever something big is happening at Adobe, not only do we invest heavily in talent and technology but also t-shirts.

Speaking of talent, let’s welcome our guests.

The first up, we have Product Marketing Manager of Adobe Experience Platform. Welcome to Hung Vu.

Hung’s fun fact, as you may have seen in the intro, is that she has recently become a new dog mom. Tell us about your dog. What kind of dog do you have? Tell us something adorable about your puppy.

Yeah, of course.

So hi everyone, my name is Hu. I’m a Product Marketing Manager on the Adobe Experience Platform team and I lead the go-to market activity for Adobe Experience Platform Agent Orchestrator. So very excited to be here to share with you a little bit more about what we have been cooking and how to work for. So as Daniel mentioned, I have recently become a dog mom. So I’m a dog person all my life. I also love cats but I always kind of identify more as a dog person. But I think throughout my whole life, I’ve never been able to have a dog of my own. We have a family dog, which I love and adore but never really feel like my dog. But now that I’ve entered my new era, I feel ready for a dog and my partner and I feel ready for a dog. So we finally signed up to get a puppy and welcome our work into our life. So my puppy is a nine month old Corgi. Corgi is also my favorite breed. He’s also a dog person but he let me have my favorite breed first. She’s super energetic, very adorable, very true blue Corgi. So needs a lot of physical and mental stimulation all the time but very affectionate. Extremely sassy as well. So we’re still trying to deal with her tantrum every day but she’s very affectionate. She loves to walk, loves running, love us. I love people in general. She’s not much of a dog. She’s more of a person dog, if that makes sense. What’s easier to train, a dog or an AI agent? Oh, I feel like I would like to say it’s AI agent because the agent does retain the memory of what has been trained for. I feel like my dog literally, she’s picked up tricks really quickly and then literally two seconds later she’ll forget and she’s like, what are we doing again? And I’m like, oh, yeah. Nice. Well, let’s welcome our second guest, product manager for Agent Orchestrator, Namita Krishnan.

Her fun fact is every job she’s had, she could see her apartment right from her desk. Now, Namita, I feel like there’s something Hitchcockian about your situation and I’m not sure if it’s more rear window where Jimmy Stewart spends all day just looking out his window at the dramas unfolding in this very small area of his courtyard or if it’s more of a psycho situation where Norman Bates’s whole world takes place between his house and the hotel he owned.

So I have you a question about Namita. Does she ever leave the meeting because mother has sent her an urgent text or has she ever tried to hang her taxidermy in the office? Trust me, I don’t think I’ve done either, but I can be. I can do it if I want to. Yeah, it’s definitely a little, not psycho, but a little crazy to think about that. I can always just see, look into my house anytime I want.

That’s good. Okay, well, our third guest, we have three guests this episode, so we’re very fortunate. Our next guest is a Senior Product Manager for Experience Cloud Interface Components.

Let’s welcome Cole Coddleley.

Hi, thanks, David. Welcome Cole. And Cole’s fun fact is he never knows which hand he’ll use for a new activity. He might write with his left hand, then throw a ball with his right, fence with his left hand, and then use a mouse with his right. Now, I think many people wouldn’t object if I said that Dexter is one of the most sinister characters on TV right now. And of course, Dexter is Latin for right-handed and sinister is Latin for left-handed, which I’m sure you knew. And now I feel, Cole, like we’re just going to have to watch what you do with your hands this whole episode. And specifically, which hand do you prefer to use on prompt AI assistant? That’s a great question. I think it’s my left hand, to be honest. It’s the sinister one. Do you feel it gives better answers when you use your left hand? I’ll have to do an annotation study on that for you and get back to you with some more data.

Just like an A.B. testing.

OK, so we’ve talked about Norman Bates and Dexter already in this episode. Clearly, it’s almost Halloween time. Does anyone want to share what they’re doing for Halloween this year or share one of their proudest Halloween costumes of years past? Namaed, do you want to go first? I don’t have plans yet for this year. I know it’s next Friday, which is so soon. But a few years back, I did dress up as Wordle, the game, the New York Times game. So I had all of the clues on my outfit. And I think, I mean, no spoilers, but I think the final reveal was either trick or treat. I don’t remember which, but that’s what the final word was. So it was a game for me and for everyone to play. I like that. I like that. Cole? Yeah, so my birthday is actually the day after Halloween, and I have a close friend who has the same birthday. So we usually do joint costumes and do a little party. In the past, we’ve done the cast of Schitt’s Creek.

And we usually have fun.

No plans this year yet, though. But that’s fun costumes in the past. Oh, cool. And Hung, you mentioned that you’re going to a really interesting place this Halloween. It sounded very creepy.

Yeah, so for those of the viewers who are joining us from the Bay Area, South Bay, you know, of Filoli, like the flower garden, very pretty in the daylight. But this year, they’re doing like a special Garden in the Dark event on October 31. I mean, my boyfriend signed us up without really knowing what’s in place.

Promised trolls, shadows in the woods and things like that. So I’m worried, but we already paid. So I’m going to go ahead and see what awaits us in the woods. Hopefully not too much because I’m a little bit of a scaredy cat myself. But I love watching horror movies, which I think amazes everyone to know.

That sounds great. Well, my family is very big into Halloween. And so we go nuts on costumes and just share a few of my favorite Halloween costumes from years past and then we’ll get into the episode. So this one I loved on the left. I went as the Baba Duke from that amazing horror movie during COVID. I let my hair grow long and I got all into the show Vikings. So that’s my Viking costume. And one year I went as a unicorn.

All right, let’s get on with the show. So let’s get started. Hong, you want to give us a quick overview of Agent Orchestrator? Yeah, of course. So let’s get without further ado, let’s get straight into it. So again, everyone, thank you so much for joining us today. We’d love to give you a quick overview of Adobe Experience Platform Agent Orchestrator. But before I go into that, I just want to kind of take a step back and think about how major technological advancement have really sparked generational shift in customer experiences. Of course, it’s changed the way we work, it changed the way we play, but more importantly for brands and for experience makers like yourself who are joining us on the show today, it really have sparked really amazing, amazing shift in customer experiences. You know, each era really brings with it major, major changes. So from the rise of the internet to cloud, to mobile, to social, every wave have transformed how brands connect with customers. And then we enter the era of AI with in and of itself is just amazing innovations all around. So we have predictive AI that brought greater intelligence to interactions. We have generative AI that unlock creativity at scale for experience makers. And now, agentic AI, which I hope you haven’t been living under, Rob, agentic AI is everywhere right now. It’s truly the next frontier of technology. It brings with itself autonomous, very context-aware agents that can orchestrate experiences with never before seen precision and adaptability.

And agentic AI has already reshaped marketing workflows at their core. By 2029, over half of Fortune 500 companies will deploy AI-driven experience agents that can deliver autonomous and personalized customer journeys. We’ve talked to our customers and we know that marketing teams that embrace these agents will automate engagement at scale. They can gain that decisive edge over the competitors, while those that don’t adopt will fall behind in speed, in efficiency, and in relevance.

And specifically, agents will continue to transform the way that marketing and customers experience teamwork. As agents take on more of that routine, more of those repetitive tasks, humans will now be able to expand our capacity. We can revolve towards more strategic orchestration, creativity, ethical oversight. We’re no longer burdened by those necessary, but very routine and very repetitive tasks. And with that, workflows will be a lot more streamlined. The tools the marketers rely on will become a lot more interactive, a lot more deeply integrated. You see that coming alive in NAMITA’s ENCOS demo in just a little bit, but you can see how AI assistant agent orchestrator has evolved to be a lot more interactive. You’ll be able to have that two-way conversations to continuously improve on your workflows. And at the same time, a genetic platform will continue to defy customer engagement. It enables brands to create richer and more dynamic experience that truly, truly drive measurable business outcomes.

And if you have joined us at Summit this year, which I hope to have seen you around, you might have seen this slide before. It appears in a couple of our keynotes, a couple of our sessions related to AI. I always find it to be such a great one to start any conversation related to agent orchestrator because it does a great job of outlining Adobe customer’s experience orchestration vision. We bring together the content piece, the data piece, the journey piece to help brands deliver personalized experiences at scale.

And the genetic AI, specifically Adobe Experience Platform Agent Orchestrator and our army of experience platform agents, the 12 purpose-built experience platform agents that you see on the screen right there, are truly the ultimate key to unlocking this ambitious vision. And some of these agents that you have on the slide hung, so some of these are available now and others are ones that Adobe is launching in the near future, right? Exactly. Yes. Thank you for jumping in, Daniel. Yes. So I believe we have around four to five of these agents already live and ready for licensed customers of experience cloud applications to start using. So we have Data Insights Agent that’s already live, Product Support Agent, Audience Agent and Journey Agent was also just live last month. So that was super, super exciting. We have many more in the works. Some of them might be available in current experience cloud application and some might be available in AI first application. So we do have a page that’s dedicated to Agent Orchestrator and these agents that I can drop in the chat in just a little bit if you want to find out more about which agent is available in which application.

One of these will be in the demo today, so you’ll actually see Audience, Journey, Experimentation, all of it. And we can call them out as well. Just so that it’s clear. And we’ll be looking for episodes that go deep into individual agents.

I just wanted to remind viewers quickly that if you have any questions, go ahead and put them in the comments of the episode. I see Barbara announced one of her favorite costumes of Halloween’s past. Oh, no. Or maybe you were just excited to let that on.

So thank you, Barbara, for chatting in the comments.

And please put any comments or questions you have in there and we’ll address them during the episode.

Perfect. I think Barbara was referring to answering your questions about whether DoF or AI is easier to train. She has confirmed that a DoF is easier to train. I feel like I need some training tips on Barbara because my process is not- I wonder if Barbara, maybe you can open an AI and ask it whether it’s easier to train itself or a corgi.

All right.

All right. I’m back to you. Perfect. Yes.

All right. So for those of you who have never heard of Agent Orchestrator before, again, this is a great definition for all of us to be on the same page. So at Adobe, we think of Agent Orchestrator as the central technology for building and coordinating AI agents in Experience Cloud is truly what enabled them to perform very complex decision-making and problem-solving tasks. So it is our centralized and connected approach to delivering high-quality, purpose-built, very extensible agents. And as the name suggests, it is also an orchestrator master. It orchestrates multiple agents to deliver measurable results and support end-to-end customers experience use cases.

And when we build Agent Orchestrator, we really desire to serve multiple audiences. Of course, there are practitioners like yourself who are joining us today. We designed it so that it can empower teams with intelligent agents that can boost productivity. It can automate a lot of those routine and very repetitive work, and it also surface insights that really improve decisions. And then for the customers that you serve, for the end customers, Agent Orchestrator can enable branded agents that engage with your customers in real time. So it adapts to the customer intent, adapts to the customer context to deliver that truly personalized interaction that, as you know, is very important in today’s day and age. And then last but not least, for the broader ecosystem, Agent Orchestrator opens the doors to interoperability. It empowers and allows Adobe agents to collaborate with third-party agents, customer agents, and platforms so brands can truly tailor, can truly customize and configure and extend those experiences in a way that meet their unique business needs. Yeah, you know, Hung, I’m going to pause you there for a second on that last section about the ecosystem.

Daniel in the chat says that he’s curious about cross-platform orchestration. When will the Agent Orchestrator be able to hand off tasks to third-party agents like Salesforce or ServiceNow agents? Yeah, and for this, I don’t know if Namita or Cole, if you have a closer look into this than I am, yeah. Yeah, I can answer this, right? And definitely, Daniel, that’s a very good question, right? So from day one, we are building the platform, the Agent Orchestrator to be open and extensible because we do realize that everyone in the industry, whether it’s our customers, partners, everyone’s building their own agents, and this can only work if we are truly able to do some cross-platform orchestration. So as of today, we don’t have that capability to actually integrate with third-party agents, but it is coming very soon. And we’d love to work with you, hear from you on those specific agents if you’re interested in either in your organization or a third party so that we can really tailor the work we’re doing to enable those use cases.

Just to add, I think hearing about the agents, specific agents and products that you want to work with, but also specific use cases is great for us to hear. So we’d love to hear more from you and from others about the opportunities that you see, especially on the ecosystem side. Is there a preferred way for people to get involved in that process? That’s a good question. We have some feedback forms, but in the chat today is a great place to start. And then always reaching out to, if you’re an Adobe customer, reaching out to your account rep.

We help with community forums. There’s a couple of different places, even in product feedback is great as well.

Great.

Awesome.

Sounds good.

The demo is coming. I’m not just going to be talking for the next 30 minutes. So a couple of different ways that our Adobe approach is uniquely beneficial to customers. So first and foremost, it’s grounded in real time customer experience orchestration data. So our authentic offerings are built on really rich customer’s experience data. It’s not just profiles or audiences, but also content, also operational data. So this really strong foundation give our agents very deep semantic understanding of customer contacts, making them really a lot more powerful than solutions built on generic and data sources. And secondly, and also thirdly, agent orchestrator is purpose built for experience orchestration and also extensible for enterprise skills. So we’re not just creating a very general purpose, agenda platform, right? Instead, we focus on really what we do best, which is orchestrating customer experiences. So our experience platform agents are built to excel in this domain while still remaining extensible. So it’s able to integrate with other agenda platform and orchestrate alongside external agents. And then last, but definitely not least, responsible AI is truly one of our differentiator. So we always lead with trust and have always led with trust, you know, from Firefly to our content authenticity initiative to privacy first principles in experience platform. We have always set a strong standard for responsible AI. So we’re sending these trust protocols into our agenda framework and making ethical AI a core differentiator for Adobe.

And then as we think about customer experience lifecycle, I’m bringing back the top Adobe experience platform agents that you’ve seen in the previous slide here. But as you can see, you know, every stage of the customer experience lifecycle, we want to be able to have Adobe experience platform agents there to supplement, to support, to even take on some of the tasks of marketing and customers experience teams so that you can, you know, unblock any of the roadblocks. You can expand the capacity and you can truly, you know, channel your energy and your talent into areas that are more uniquely human, strategic orchestration, creativity, and many more. So these experience platform agents are purpose-built with very deep customer experience intelligence, and they can draw on Adobe’s expertise and leadership in delivering personalized customer experiences at scale.

So I think that’s all from me. You’ve seen a couple of these here and they will come to life in let me does it calls them all in just a little bit. So over to you, Cole.

All right. I should be sharing. Hopefully you can see my screen okay.

I’m going to dive into the demo. We have a lot of prompts that we want to get through today. So I’m going to be driving, but I hope all of my, my co-guests will, will jump in as well. I’m on the landing page right now, and I have this new home widget where I can start asking AI assisting questions. I also have this left nav option, but I’m just going to go ahead and get started and maybe greet AI assistant with a good morning. My, you know, one of the biggest innovations we have here is the ability to have emojis in this chat now. That’s very important. But let me tell you a little bit more about what we can do. So one of the cool things about agent orchestrator one thing that I really like is that you can always go in and start asking questions about what it can do rather than just having to figure it out on your own. So let me add a question here related to that. What can you help me with in terms of understanding? Is this a situation where you’re using both hands to prompt AI assistant? Fortunately agent orchestrator can deal with my typos, which is, you can see here, I got a response so I can see, Oh, well agent orchestrator can help me with data exploration insights. So some stuff related to data insights, agent, some audience management things as well. And it gives me some sample prompts for these. And then also customer journey analysis. So I can ask questions about journeys here and start to understand my customer data that way. I’m going to start with maybe something related to revenue. Yeah. Daniel, go ahead. One question that came in about, you know, capabilities of AI assistant and the prompting. There was a question about language support. Does the AI assistant support multiple languages? Will we be supporting multiple languages in the future? Yeah. So right now we only support English, but that’s on our roadmap to support additional languages, just like we support in, you know, experience league as well. So something that we want to work towards, if you ask questions in a foreign language, sometimes you’ll still get a good response and you can operate it that way, but fully supported.

That’s still on the roadmap for us to do.

I’m going to jump in with another question here, just to make sure that we get time to show all the different agents that we have available today. So now I’m asking a question related to revenue and we’re going to get a reasoning block here so I can see that we’re actually querying information related to customer journey analytics. As agent orchestrator is thinking, I can always click into this block and see some of the reasoning text that comes back. And here you can see a chart. So I actually got a nice chart in here. I can see revenue over time which is great. If I wanted to change the view here, I could make this a table and I’m on a little bit of a small screen right now, but I could also expand this to my right-hand panel. There’s a split view here as well.

So some great options in terms of visualizations here. Interesting. Cool. Could you go back up and tell us a little bit more about the reasoning step, what’s going on in there? Yeah, I’m actually going to see if Namita wants to answer this because Namita is very close to the reasoning work. Maybe Namita you can just move around there. This is something I absolutely love because I’m sure folks have used many of the other reasoning models out there. You get to see how the LLM or the agents are processing your input, what are they doing, what’s happening. And this is just a good way to make sure that the agent has understood your prompt. It’s interpreting it correctly. So in this case, Cole asked for some kind of revenue data and you can see how the agent is interpreting that request. So it’s understanding that it’s business metrics. It needs to create a visualization. That’s an assumption it’s making based on Cole’s request. So you can kind of see it’s an almost internal chain of thought as it’s going. And you can also see the actions it’s taking. So the way we’ve represented it is the icons that you see. There’s like a different icon for the two sentences. And that’s basically showing what’s the thinking happening and then what are the actions it’s doing. So it’s initially thinking and understanding Cole’s request. And then based on that, it’s performing an action, which is in this case, it’s gathering insights. So you can see kind of both how it’s thinking things and then calling different tools or different agents. In this case, it’s actually using the data insights agents. So it’s actually invoking that agent to get us this beautiful bar blank chart. Yeah. I like that it’s showing which application is pulling the data from, in this case, CJA, because I mean, it seems like an orchestrator, it’s at the experience cloud level and has the potential to access stuff from our many, many experience cloud applications.

So being able to confirm which one is being used as the source of this information, I think is really cool. And yeah, I mean, that’s definitely a big value prop for this experience that you don’t have to keep switching between CJA and RTCDP and all of the different apps. With the right context, it’s able to understand which data source to fetch it from. Of course, you might have to try a couple of times in a few cases to make sure it’s understanding and going to the right place, but you can do all of that in the same interface. Great call outs on the reasoning side. And you’ll see that in a bit too, as we transition from some questions related to CJA, there’ll be a seamless transition to stuff related to RTCDP. Let me ask another question here. Let me show me revenue and we’ll say by city in the same period. So we’re asking a question that relates to the previous topic. I’m not actually specifying the period here. I’m just saying, show me revenue by the same period. And you’ll see that aging orchestrator is great at being able to capture that. Shows me the same period of time. And I get this nice bar chart so I can see all my cities. And when I expand this out too, I can actually see the full list of cities here in table form, which is exciting, but also can be a little bit difficult to read because there’s so many. One thing I like to be able to do here is I can change this chart type. And if I select number, I can actually see the total number for this particular query, the total revenue across all these values, which is when for me to be able to do.

And someone asked, Cole, I don’t know if you have any prompts ready to go to be related to churn analysis. I’ll say that one more time. Turn analysis. Churn analysis. Oh yeah. I don’t have anything ready for that today, specifically in the context of like CJA perhaps and seeing information on that or information related to journeys. We do have some fallout analysis on journeys, which we’ll get to, to look at where people are being lost over the course of the journey. So hopefully that will be something that’ll be interesting for that viewer. And we can get into that in a bit too. And one question with your, so you’re asking CJ, you’re getting all of this data from CJA. How do you know which data view is being used for those questions? That’s a great question. So you can see in the title here that we actually call out the data view and the timeframe in some details, because this is at the experience cloud level, I can actually decide which data view and which sandbox or other tenant types that I want to use with each query. And I could bring them in here if I wanted to change them. So with this little bar at the bottom, I can make those changes. I could select a different data view here and I wouldn’t be lost for my conversation. I could just keep my conversation with some new data. I can also do the same thing for sandbox here. If I wanted to look across another sandbox, really quite powerful when you’re using this, either in the full screen, you still have this experience in the rail view too. So if you wanted to ask questions about your sandbox while you’re trying to create something in prod, you could do that as well. But all of that context setting is right here and it’s always visible on responses too. So depending on the response, when we ask some other questions related to like data and AEP, you’ll still see some of that information on the explanation. But yeah, that’s the overview of the context there in the data view. I want to just, maybe I’ll ask it to limit this real quick too, just to show you what that looks like. If you want to ask questions that aren’t completely a data question, but just a continuation of your prior prompt, you can always do that and see some adjustments made to the visualizations that you’ve already created.

And I know that we were talking about maybe looking across a couple different use cases. So I’m going to jump in and show you a little bit about what it looks like when we move to data that might not be in CJA. So let me ask a question related to that.

So keeping with our city theme, we can now go to audiences and AEP and see, do we have any fields that also help me identify city in my platform products? Oh, so trying to find out which XDM fields the implementation uses to capture the city is what you’re doing. Exactly. And I’m really just asking, do you have any fields that would help me identify city? I’m looking through all those fields and giving me some recommendations about the relevance of the fields that I have. So back with two here, home address city and billing address city. And I can see most of the context here, but I can always click show more or expand and open up that right panel and see the full list of, for example, the usage contexts. You can see where this field is actually being used across my other objects in the platform. Yeah. I mean, this looks great for a marketer, business user who they need to know what XDM fields do what, but this is the kind of stuff that usually only the implementer, the data engineer is going to have off the top of their head. So instead of having to open up a schema and drill into the fields and do all that kind of regular role, this seems like a great quick way seems like a nice performance enhancer. Yeah. You don’t have to know the exact name of the fields, which is great. You can just go in and ask a general question and then get some results related to your search. And it doesn’t only just work for XDM fields. If I start asking questions about audiences too, shall we audiences that use let’s just say home address dot city. I should be able to get audiences as well. I want to see results related to that.

One thing you might see here is the streaming in of the responses. So as you know, something coming in, it kind of tries to stream in that response in the UI so that you’re not just looking at a blank screen for a minute. So that’s really nice. Yeah. It adjusts the responses after it comes in too. So sometimes you can see it’ll truncate things and put them in sections where it’s like, Oh, you might want to know this, but I don’t necessarily have to show this to you because it’s not exactly what you asked for. But if you wanted to dig in a little bit deeper and see something like the sequel or how the results were gotten in text, you can do that. And for this one, I can download it too. I click download. I can get this information. Yeah. And so I’m going to pause it here for a second. Cool. Because you’re showing all of this cool stuff. And I know people who are watching are wondering, well, how can I do all this cool stuff in my own org with my own data? So we have one question. What are the minimum licensing requirements to get Agent Orchestrator in the product? And how can they start using it if they have RTCDP, AJO or CJA? What do people need to know to get started with this? Yeah. So there is a SKU that’s available. And now it’s a feel free to jump with more details here. Or Kwang, you might be the best person to answer this, but there’s a SKU available that people can get a $0 SKU. And then if you have any of these products, you don’t have to have all of them. But if you have the products that are currently supported, RTCDP, AJO, CJA, then it can be available and turned on for you. Yeah. I can jump in quickly. So if you’re already an Adobe products and you’re already owning some of the experience cloud applications, which sounds like you have maybe CDP, AJO, CJA, then you can get the $0 promo SKU that we have going on right now. And you’d be able to access Agent Orchestrator. So also just to clarify, Agent Orchestrator is not a product that you can buy. It’s an orchestrator that orchestrates all of the agents that come with your license experience cloud applications. And also just to clarify, not all of the 12 agents will be available to you. It depends on the applications that you license. So for example, if you own CJA, you get access to data insights agent and products support agent. For audience agent and journey agent, you need to get CDP and AJO. So I should be able to drop that link with the available agents and where and in which experience cloud applications they still face in. So you can refer to that. So if we have practitioners watching the episode, so they need to find the person at their company who is speaking with Adobe and contract getting the licenses and let that person know that they want access to Agent Orchestrator. And then that person can work with the Adobe rep to get the license and get this provisioned into their account. Right? Exactly. Yes. There’s also an additional permission step, I believe. But that permission will also kind of correspond to the permission you already currently have in your license experience cloud products. So if you already are able to create audiences in CDP, you’ll be able to use the audience agent to create audiences. Okay. And then there was one other question that’s sort of in this area about getting started with Agent Orchestrator and AI assistant.

And the question was, how will customers be able to audit and verify the ethical AI component and ensure compliance with their internal AI protocols and standards? I think that’s where our fact sheet comes in probably.

Yes. That is a great and very timely question actually, because we just recently published our Adobe experience platform Agent Orchestrator and Agent Security fact sheet. So that has all of the key information about user authenticate, like the usual security questions that we receive from customers. So user authentication and authorization, data retention, data processing and storage, LRM services. So that is a great document to get started with, to go through and get answers to some of your common questions. And then if your internal AI team still have further questions, we’d be happy to work with your account team to answer those questions, jump on a call with you and make sure that we address any questions or concerns that you might have about using our products. Great. So once we wrap this episode, can we put the link to the fact sheet in the YouTube comments? And then if your questions aren’t answered in there, it sounds like the next step is to work with the Adobe account team to get any additional questions answered through that relationship. Great. Cole, back to you. Sounds good. Just to add one point to that too, Hwang mentioned this, but when we were building Agent Orchestrator, one of the things that we want to make sure it does is not extend anything, any permissions beyond what your user permissions are. So when you ask questions in Agent Orchestrator, who you are as a user is what gets considered and your permissions get considered with each response. So you won’t see a lot of limitations on this demo because I have access to in this environment, but whenever you ask a question, Agent Orchestrator respects you as a user and doesn’t extend and answer questions beyond what it can answer, what you can do in the product yourself. So an example of that, Cole, would be something like, if you’re asking, show me the top five audiences, you would need to have permission in either RTCDP or AJO, the permission to maybe it’s powerful to view audiences or view segments. If you want to view segments or view audiences in the Agent Orchestrator, you need to have a permission item to view audiences or segments in the application. Exactly.

If you didn’t have access to certain objects, like if there’s an audience that you didn’t have access to, then you wouldn’t see that object either. Yeah, it sounds like there might be a little bit of typing, but I’ll keep going here. I know we’re short on time and we still have a lot of prompts that I’d love to show you, but we can also do cool things related to audience in terms of seeing how much things have changed over time. I can see size changes for audiences and I can just see ones that have happened in the last seven days. I could change this around and I will actually, let me actually break one of these out so it’s a little bit more visible. For the second audience, show me the weekly trend going back three months. So there’s a lot of cool stuff we can do in terms of analyzing audiences, analyzing data that you’ve seen in CJA already, and then I’m hoping we can get to experiments and journeys as well. But just wanted to show you a few more prompts here on the audience’s side before we jump into journeys. I love that you said the second audience and it picked it up from the table. So you actually see that in the reasoning block. The second one was experience event and you kind of see that disambiguation happening.

It’s great too at the end if we have time, I’ll try to summarize this whole conversation and it’s great to see that the whole elements of the conversation can be quickly understood and organized. Super helpful when you’re just having a lot of conversations and then you come to a point where you’re like, oh what did I do or what do I want to pull out of all of this. There’s a lot of great continuity here and you can see I can do some deep analysis of different types on audience sizes. I can see this whole table for the audience that I selected here and how much it’s changed over time and week by week. I would spend more time here but I really want to keep going. I’m going to just look for my Black Friday journeys and I’m not going to specify anything besides just saying Black Friday. Now we’re looking at audiences and xtm attributes. Now we’ll start getting back some journeys and we can do some cool new analysis that you probably haven’t seen before on these journeys that we’re getting back. So I have two here, pretty straightforward response but now let’s get into some analysis of what we can show with these journeys. I’m going to start with conflict analysis and one other cool thing.

I’m just going to do the first one again since I really liked and Namata really liked how that interaction worked last time.

And by the way to address one of the questions in the comments, Cole right now is interacting specifically with these journey constructs which only exist in Adobe Journey Optimizer. So if you don’t have Journey Optimizer you wouldn’t have journeys both under this construct. So if you only had an RT CDP license these types of questions wouldn’t be relevant. The more abstract version of Journey maybe there’s stuff in CGA data that you think of as a journey, a customer journey you would be able to query that CGA data but this requires Journey Optimizer because it’s stuff that was built in that product.

Yep, I call it around a lot. So we’ve jumped from Customer Journey Analytics through RT CDP and now we’re into some AJO Adobe Journey Optimizer constructs. Just to go through some of these results, so I asked for some conflict analysis on one of these journeys and it was able to tell me both audience and schedule overlap. So it’s told me hey you have some audience overlaps for a couple of the journeys that are running at the same time and you have some scheduling conflicts and it’s done some severity rating based on those. So I have this text block which I’m not much of a text block reader but I also have this nice table where I can start seeing severity information and whether it’s schedule based or audience based. So some cool stuff there. It’s given me a little card for the journey too if I wanted to do some actions off of that I could do that as well.

So really a nice in-depth response on this query that I’ve asked a question about. And this what you have up right now, it was asked can the agent not only analyze but also suggest actions and looks like what’s happening in that recommendations section there to help resolve these conflicts. Yeah it’s able to give me some recommendations here. You’ll also see that for like the experimentation stuff. It’ll do some analysis and thinking and give you some recommendations based on what’s happened in your environment. Great question.

Let’s do a couple more things here. What else can you tell me about journeys? I just like to remind people about this one because I oftentimes get stuck or sometimes I’m just like what else can you do and agent orchestrator is great at that. Giving me a couple more things. So we did conflict analysis already but we can do some additional stuff here as well. So I can ask for like drop-offs for journey. So let’s do that really quickly too. Show me the drop-off for and then this is a fun thing. If you click this plus button or I click plus right here I can just type in I can start auto-typing and I should get some examples here of what’s available. It’s really nice for being able to cite specific objects. I know a lot of people have long object names or audience names and it’s a great way to just be able to pull in the audience that you’re thinking about if you have the name off the top of your head. I know we’re running a bit short on time so I wanted to see if we could get to maybe one experimentation question after this but I wanted to show this prompt for you too. So this is a cool new visualization and response type that we have. Again we’re getting the journey details back in terms of the journey name but then I have this nice fallout chart where I can go and see where the fallout’s happening in this specific journey and then I have some of the information captured in observations and again in recommendations. So I won’t dive into all the details here. I have a pretty big fallout rate right here and then at the top there’s a 50% cut but Agent Orchestrator is able to do all that analysis and it’s super powerful and exciting to see. Let me do one more thing here. So I’m just going to ask it about the experiments that we have. Show me the status of the experiments.

Oh go for it. Go with the typos.

I know, exactly.

We’ll get a summary here and we will see some of the experiments they’re actively running.

Again I wish we had more time to show you this but it’s really easy to kind of get these results back and then say oh can you make this nice and add this into a table and we could get a table result back rather than just the text that’s here. Maybe I’ll just wrap up. I know I’ve used a lot of my time but I’ll wrap up with just summarize this conversation for me. And just a quick call out on the previous one. It is based on your data and experiment with using Experimentation Accelerator. So that is a new product as well. So if you have Experimentation Accelerator that’s when you would be able to do some of these cool analysis on the experiments you’ve done by new insights or new things to experiment on. So all of that is kind of powered using the Experimentation Accelerator.

Yep exactly. Thank you, Namita. There’s a lot of different products and features that you’re able to access here and as we add new things to those agents that that Huang talked about, as we add new products, those things will all be accessible through Agent Orchestrator. We really want this to be something that connects to every part of Experience Cloud that you’re using and helps surface information quickly for you. And one question that came from our executive producer Mugdor, so we better answer this or he’s gonna fire us all. A follow-up question to those recommendations that you got in the answer. Can the Orchestrator actually do those recommendations for you? Or if it’s like you should create this audience, can it create that audience for you? That is a fantastic question. So creation stuff is coming very shortly. We almost thought about showing it for a bit here because we have it available in some staging places. But that’s kind of the next step for Agent Orchestrator is allowing the agent to have hands or the Agent Orchestrator excuse me to have hands and do things like create journeys or create audiences, make changes to objects that you have in the system, that sort of thing. Yeah so there’s a lot of talented people building out Agent Orchestrator and the agents and capabilities. So watch their release notes because there’ll be a lot of features I expect coming out in the future.

Yeah I’m sure we’ll be doing more of these webinars with some cool new creation stuff and acting on recommendations which will be exciting. I know we’re about at time. Thank you for letting me walk through the demo. There’s tons of other stuff we’d like to show you but hopefully we could do that in a follow-up session. Yeah and Cole these prompts that you’ve been going through it’s really helpful to just to see what you can do like through those prompts. A lot of these are published as part of the documentation for the agents right? There’s agents like the there’s the documentation and experience that you can ask questions. You can say give me 10 sample problems and you get some sample prompts there. I’m going to go back to the to the to the like a new conversation too and when you start a conversation you can see these little light bulbs or pills and you can click on them and see some example prompts related to different general areas of using AI assistant. So feel free to check those out as well. There’s a few different resources for getting some sample prompts and getting going with Agent Orchestrator.

Great yeah and a question from Daniel. He’s asking does that homepage agent do you need to select the agents at any point that you want to use? Nope that’s the the beauty of Agent Orchestrator is you ask your question and then Agent Orchestrator will decide which agent is most appropriate for your question.

Yeah and I think that’s the beauty of it as well right? You don’t have to specific and as in Cole was showing you actually don’t know which agent is performing the work. It’s more so that Agent Orchestrator kind of detects the user intention from the prompt that you put in and it will orchestrate the right agent or agents to help deliver that use case for you which I think is really user friendly as well because sometimes it’s hard to recall like what agents do I have access to. You don’t have to know that you just have to come with a goal of what you want to do in the interface and you can ask AI assistant and Agent Orchestrator and you’ll be able to perform those actions for you. Yeah and I think it’s very powerful not only for if it’s a product that you use a lot to just help you do things faster with that product but also if it’s a product that you don’t know like you know it’s like I never fully mastered the analysis workspace in place that’s in CJA so being able to get data out quickly without having to click around through that interface is huge. Well it’s time to wrap up. I think we’ve got through almost all of the questions in the chat. There might be a couple that we didn’t get to so we’re going to do a post in the experience platform community with those remaining questions and get those questions answered in that community post. So we like to end each episode with an unrelated cool tip.

This episode’s unrelated cool tip is brought to us by Senior Product Marketing Manager Hong-Bou.

Awesome thanks Daniel. So this is like a tip that I feel like maybe some folks have known but I just never know and this is related to kind of finding recipes online so I’m not much of a cook myself but whenever inspirations strike or when I survive like a cool dish on TikTok on YouTube shots I kind of want to recreate it and then I start looking for recipes on kind of just Google right and I found like a really cool one that has 4.9 star and then I start clicking on it and then it’s like a five-page life story about someone and then how this dish brought them to like the time in their summer home with their grandma and I’m like that’s great and all but I really just need to get to like the ingredients. Yeah exactly like especially when I’m already at home for I’m like okay I really want to make this dish but then I just have to keep scrolling and scrolling just to get to the ingredients to access whether like do I should I really make this recipe or is this really out of my the the the lick of my culinary skills. So one of the cool tips that I found recently is that at the beginning of almost all articles there’s going to be a print button and then if I just click on that it will basically bring me to a PDF ready to be printed PDF that just have the the essence of the recipe which is the list of ingredients and the step to step by step instructions on how to cook that dish and that’s really all we need when we go to any recipe online and I found that’s honestly kind of changed my life a little bit so I don’t waste any more of my time just scrolling through. Sometimes it can be 10 pages and I’m like this I just need this one list of ingredients. That’s awesome. Yeah I love it. Thank you everyone. Thanks guests. Thanks audience for joining us on this episode of Experience League Live.