Agentic Sites: Building Hyper Personalized Websites — Carlos Sanchez, Adobe

What an agentic site is trying to do

00:13Hello. Thank you for coming. Um, I'm going to talk to you about aentic sites. How we call it is bu building hyperpersonalized websites. I'm not going to just talk about it. I'm going to show you what we're building. Um, I've been working on on this project for for a bit now and we'll try to show you what is possible today with with AI.

00:36Uh, I work at Adobe at uh I'm a principal science at a product that not many people know, Adobe Experience Manager, content management. We run a lot of uh websites, properties for big brands and my background is in in open source uh contributing to to a lot of foundations and projects.

01:01What are agentic sites and how are we building this thing? So we are looking for sites that are uh looking at the what intent the user browsing has. What is the user doing? What are the user trying to achieve? And the end goal is to personalize these pages for the for the current user browsing so that eventually this uh drives higher engagement or uh conversions whatever the marketing teams want to want to achieve.

01:35And these pages are personalized in real time based on the on the user that is uh accessing the site and what is the what is the user doing. The stack we're using is AMH delivery. So this is the part of the product we we have uh where all the content is on the edge and then we have a backend service that powers this experience with uh uh different LLM providers uh LLM services.

02:09Uh we use cerebras for fast inference or we can use also we try bad rock and and a bunch of others. I'll be showing cerebras today. Um and you will see the reason why the the engine that is

Personalizing blocks, not whole pages

02:24personalizing this uh this bit it's is um is using the rich content and blocks. So different blocks on the site are customized depending on on what the user persona is. We don't want the the whole site to be generated. I mean if you talk to marketing people they they have a very strict brand guidelines.

02:42You don't want to just came up with or have some hallucinations there. So the what is personalized is different sections of the site and we use the whole site as a as corpus. We built a rag from the whole site. So what is generated is grounded on on the existing site.

03:07We try to solve the problem where yeah one side fits all. We want hyperpersonalized experiences. Also, we want to help uh our customers to do more automatic um authoring. So, not having to create thousands of different variations of the site, but uh use AI for this and then do

Grounding generation in the site itself

03:30this multiple layers of of personalization. uh some examples of of what we are doing or are showing the demo is a instant persona adaptation query generation when the user search for something on the site the page with the results is customized for them and also uh something like recommendations where after you browse the site for a period of time we recom we can create a page that recommends something based on on on what you are what we think you are looking for marketeers uh they can define this strategy on natural language and they can use analytics to to drive the loop of personalization

04:13and what is the end goal and how this goes back again to change to adapt the personalization to improve that uh whole cycle. Everybody's talking about loops in this conference. So that's that's one of the loops there. How the architecture look like?

04:30So it's a dynamic front end with some blocks what I mentioned before and with edge

The architecture behind the blocks

04:37delivery services is basically you compose these blocks and uh they are updated on on real time through with the AI the back end uh we we do the evaluation of the models and the providers and one thing we realize is that this is very dependent on the site.

05:00So we have a bunch of prompts and we look uh we run it across a huge variety of uh models and providers and then we look at the accuracy we look at the speed but this is going to depend highly on what type of site like how big is the site how I don't know what different u different um area is the site uh targeting what what type of commerce it is and so on.

05:23So we we run this this um evaluation continuously. We use uh prompu. Uh anybody heard about prompu? Okay, some people. So, PROMFU allows you to evaluate models and prompts against multiple models providers and you can do local models and any of the a bunch of

Evaluating models per site, for speed as well as accuracy

05:50open compatible uh providers and and a lot of them. Basically, we look for two things, right? Accuracy. That's that's typically what people look for. But also we want the speed because we don't want the site generation to take more than one or two seconds.

06:12Right? Because people um this is already proven that people want the the faster the site the more conversions it generates or the better the experience it is for the user. Um yeah what I mentioned is different sites may have different requirements.

06:28So you may have to run this uh evaluation of models depending on the site. This is an ex uh we we executed this some of these queries. So we have uh 15 prompts for this example site and we have uh at the top you can see with cerebras on the gemma 4 model that was announced last last week we can get an average latency of 1.1 seconds generating a p page you you can compare

1.1 seconds against 4.6

07:02that to the second one which is 4.6 in six seconds, right? So the difference is huge and that's why uh we use Cerebras for for this use case and uh you can see that different providers different models have different um different speeds and here is uh let me I can show you the whole thing here.

07:25Not this one, this one. Right. So at the at the bottom we have other other that's sometimes uh maybe some of them may be good. They don't need to be perfect but they're good enough if they're fast enough. So that's going to be the the kind of decisions that you need to make on whether the model is good enough for your use case or not.

07:53Yeah, we're looking yeah average 1.1 seconds and then the the next ones are going from 4 seconds higher. And you don't need a huge LLM to do this sort of work because you are generating text. Uh you are deciding where to put blocks and

Why this does not need a frontier model

08:10how to organize the website. You don't need lots of information for that. So this browsing and the queries uh is are being recorded. So these are the metrics or the um the the data we gather from the user and this is feed into the LLM to personalize the site and then in this example we personalize the hero card, the products, the blog feeds and and the navigation based based on the persona also.

08:39What are some of the buttons like our call to action navigation? You can also we can also personalize those. We uh we create and I'll show you the a for you page which is a recommendation. And this is a interesting one because this you could uh pre-generate right as the user browses your site you gather these signals and you could keep generating this.

09:07So in this case you wouldn't need so such a big speed but um

Pre generating a page before it is asked for

09:16but that's interesting because it it would be if a user wanted to buy something you could just say okay for you we recommend these three products or or something like that. Um yeah and then they can see this recommendation and if they go there that that could be prefetch for them and obviously you have to keep updating it as the user navigates around the site and and so on.

09:39So that that's also something to consider on the cost cost uh implications of doing multiple generations multiple LLM calls. When uh when the user runs a query, a dynamic personalized page is shown to them. Uh when the these queries are also grouped into personas or intent types.

10:08So what is this guy? So what is this guy trying to do in the site is trying to buy something. He's trying to just get information. So you can get marketers to decide what type of groups, how many groups you want to have, how you want to

Letting marketers define the personas

10:23deal with with customers and AI will choose the the blocks and the suggestions for for those groups of people. Um and we can adapt yes the the different blocks the the the sequence of the blocks and uh media. You could also do media. One of the things we consider is uh there was some a model announced uh today or yesterday the the nano banana light.

10:48So you could even generate images very fast on the fly. Obviously not as fast as test but uh that's also uh something that would be I don't I don't know if is that something like marketing people would want to have generated images that depends on on the quality a lot if it's on brand and the site uh in this example we have a a product site and then we have guides experiences blogs and the whole response of the LLM is grounded there and uh there's comparisons.

11:21We can do comparisons between products that are tailored and the product pages can be tailored for the for the user. Okay, this this is a bit of the stack. Um not going to spend too much time here but the browser you have some layers you have the browser where the signals get um get uh from are got are got from the from the user and then we have the back end.

11:52Uh we can have the back end we run this some of this in in Google some of these are Cloudflare. So the back end is basically just calling the LLM and doing some reasoning using the rack that is built on on the site to do the generation and you have obviously you have to have the vector database the inference uh machinery and uh the experience manager is doing the serving the the at the edge is serving the the pages and the static content.

12:25So let me show you because I think this is uh so we call this uh audience of one because the idea of in marketing uh they they always

Audience of one

12:44dream on being able to personalize things for each individual. So we call it yeah audience of one. So I have this this site uh this is a site that is absolutely generated uh example site is a coffee uh machinery. So, I can go and and read some stories and I can go and look at some products.

13:02Let's go and look at this product. I can spend some time here. Uh let's go and click here. Okay. So, I'm look I'm browsing around the site and I have this debugging tool thing uh which uh let me go here. I think let's see. So, down there is the signals that the that the browsing is giving us.

13:40So I don't know if you can see it much because I cannot see it much. Then so the user is bucketed into the exploring

Live demo: signals, buckets and a For You page

13:55category. We have the pages that have uh have visited and then we have how much time is spending on each page. All of this data is now available for the LLM. So if I go here, I already have a for you page that was generated for me. and uh based on my browser and you will not notice that it's slightly different than everything else.

14:26But if I go here and I run a query uh like I want I am looking for a coffee machine to uh prepare coffee while camping. the site is this was just generated for me. And then you're gonna see some things like the text is customized camping shouldn't mean compromising on your uh whatever routine uh the coffee tips for camping and machine that are being recommended and or the nano which are um good for um

Asking for a camping coffee machine

15:09for the for a camping trip, right? So you saw how fast this was. I'm going to run it here. Uh something similar that I had here. And I can run it on the debug mode here. And you will see let's make this bigger. Total time 164 seconds to generate the page.

15:30So this includes the round trip to the LLM. This is using Cerebras Gemma 4. So the the Gemma model from Google running on Cerebras on their uh very fast chips uh we get 2,300 tokens per second which is not bad I would say. And if I run it again uh probably something like that uh the LLM time is 1 second and again 2,200 tokens per second.

16:03This is something that we only dreamed about before. On the on this site, example site, we have some other options. Uh so because we've we've been showing this to customers. So we have the the ability to change uh the different models temper temperature tokens and so

Swapping models on the fly

16:25on. And we can uh we can show uh and try the different models and see how they behave. Besides automatic test with prompu then we can uh manually come and and tweak things and see and see how that how that works. And uh we also have uh of one labs.

16:46So we have we built this tool that generates an agentic site for any site we want. So if somebody wants to have a a demo for a customer, come here, enter the URL in less than an hour. You have an agentic site. I did this last week with the AI engineering site.

17:02And I got this site that is just a search box and a few things. And uh let me open it here. The full page, not this one. Yeah. Okay. So I could say uh Europe AI conferences. So these suggestions are also AI generated and I get a page that is uh more focused on

Turning any URL into an agentic site

17:34it should be more focused on on the on this European conferences. If I go back did I go I can search for anything the same way I did with with the ARCO. I say Pacific. There was someone that was generating a good comparison side to side. Let me see if this one.

17:50Okay, here this one. I went and this generated a page with uh very good comparison. If I'm looking at two conferences and I need to decide if I figure out that the user wants to do that, this is great because that gives them a side-by-side comparison on the fly.

18:14Now this this is I think this is cool already. But then we have uh I have this idea that probably the um a bunch of people are we are talking about is the web dead? Is is the web the future still and so on? Nobody knows. But we can also do something with this uh with this audience of one these generative sites.

Where this goes next

18:42So imagine you have uh you have your personal assistant and you ask a query through in this case through Google and you say I want to buy I don't remember what the query said was something like I want to buy uh a machine and I get this on my Google TV right so this is absolutely personalized to my query okay no go back this is absolutely personalized to my query so I'm there in my living room.

19:04I don't need a phone. I don't need a computer. I don't need anything. Just my voice and something that will uh kind of show me something that is absolutely personalized to to me. Okay. So, that one. So, what I was trying to show and hopefully you remember from this session is that this is now possible.

19:38It's only going to get better from here on. It's only going to get cheaper. It's only going to get faster and you will be able to have uh huge personalization options for sites and for other things. And you can do this uh with intent driven.

19:52So what is the what is my user trying to do? What does my user want to buy? These sort of questions. And you can uh assemble a page just for them. And you can also do this with uh multiple models and and eventually it's just going to be faster and faster, right?

20:19So that's it and thank you for coming and I hope you you got the idea. Thanks