Chris Bryce: All right, we are live. It's Thursday afternoon at noon Eastern time. Welcome to episode 36 of Beyond the CMS, where we have deep dive conversations with content management industry experts and CMS platform leadership. We're live on LinkedIn, where we encourage all of our listeners to ask us anything. It's a great time to have a chat about CMS content platforms and what's going on in the industry, in the space, for anyone whose lives revolve around digital content.

Chris Bryce: Ask us anything, and if you don't catch the whole show here and now, we put our archives on YouTube and anywhere that you get your podcasts, Apple or Spotify. My name's Chris Bryce. I'm the CEO of Dotfusion. I specialize in coaching customers towards the right CMS platform selection before handing off to our awesome team to design and develop the final products. We're non-sponsored, pretty grassroots. I do want to thank everyone who's made a bit of time around their lunch hour in North America and dinner in Europe for joining us.

Chris Bryce: Let's get started and kick things off here. I want to introduce our special guest, William Borg Barthet, CMS Practice Lead at Bloomreach. William, welcome.

William Borg Barthet: Thank you very much, Chris, for joining us.

Chris Bryce: Do you want to just maybe introduce yourself and we'll get some great discussion started?

William Borg Barthet: Yeah, absolutely. So as you mentioned, I'm the practice lead at Bloomreach. If you don't know Bloomreach, we're a Silicon Valley based company. We have a customer data platform and marketing automation platform. We also have a commerce search engine, and of course CMS or DXP, which is my personal expertise.

William Borg Barthet: I've been working at Bloomreach and in the CMS industry for the last 15 years, and I'm a Java architect by training. It's great to meet you and great to be on here.

Chris Bryce: Okay, super cool. So, lots of topics to cover. I want to make sure everyone gets to know all about Bloomreach itself, some of the thinking behind it, how it's different, and the market it serves.

Chris Bryce: Just before we get started, thanks to everyone for joining here. If you do have any questions, you can type them into the LinkedIn comments, and we have a bit of tooling here that will help us see what those questions are, and we'll do our best to get those answered. Typically we're speaking to kind of CMOs and some developers here and there.

Chris Bryce: Just to create a little bit of context there, but maybe just tell us about the platform in general and the market you serve.

William Borg Barthet: Sure. We're very much on the larger end of enterprise. We've got a highly customizable Java based CMS, which is powered by the JCR, that's the Java Content Repository.

William Borg Barthet: The stack would be very similar to Adobe Experience Manager or Magnolia CMS. So this is very much for the larger end of retail, large banks, governments, and that sort of customer.

Chris Bryce: Okay, cool. Well, people are going to want to know about AI, and I know you're doing lots of great stuff in that area. Do you want to kind of start a little bit about...

William Borg Barthet: Yeah, absolutely. So AI, of course, has been around for a long time, but it's only in the last couple of years that it's really captured the imagination of everyone, when they've put these AIs together to compete with each other, to generate very high quality content in terms of human readable content and understanding human prompts at a level we really haven't seen before.

William Borg Barthet: And this has been integrated very rapidly into a whole suite of tools, and CMS is no exception. We've seen pretty much every CMS vendor put AI capabilities into their platform. Now your mileage will vary very much with these tools. In some cases they've really bolted it on there, and it's really not too different from having ChatGPT open in another tab.

William Borg Barthet: There's other cases where they've done a much deeper integration, and the real key to successful AI integration in the CMS, in my opinion, is all about context setting, if you perhaps would allow me to elaborate. If you go to ChatGPT in another tab, you're going to have to set all of the context yourself, but with the prompt in the CMS, if it's done correctly, the AI, the large language model, is loaded with the context of where you are in the CMS.

William Borg Barthet: If you're looking at a particular document or a particular page in the CMS preview, that context is given to the AI so that it knows what content you're looking at. But beyond this, some vendors would bundle their LLM with their CMS offering. Other vendors, like ourselves, we're not really bundling the AI, but rather creating an AI integration there.

William Borg Barthet: And the idea is that as a large enterprise, you're going to want to curate your own LLM, and that LLM is going to be validated by your InfoSec team. And it's going to have the context not only of where you are in the CMS application, but also of all your organizational context. So with an agent-style chat bot in the CMS, you could do all kinds of use cases these days.

William Borg Barthet: Like, tell me about the spelling, the sense, translate this into Swahili, but with the organizational context, you could also ask things like, does this content conform to our legal guidelines? Does this HTML match our design system? And then the response is much deeper than you would get from a simple copy paste from ChatGPT, and then there are the governance aspects on top of that.

William Borg Barthet: The developers and maintainers of the integration can put all kinds of guardrails to ensure that there's strict usage limitations on the chat bot inside the CMS, that there's perhaps some hardcoded rules on what responses are allowed, preventing certain errors or misconceptions.

William Borg Barthet: And it's pulling from data, not just from your CMS content repository, but also from other areas of your business. Perhaps it's connected to your customer data platform. Perhaps it's also connected to your ticketing system for service tickets. It could also be connected to your Confluence for all kinds of internal documentation.

William Borg Barthet: So what we're perhaps seeing is we're moving towards a world where companies and large enterprises procure an LLM, curate it, and then connect their various applications to it, rather than having sort of baked in AI agents inside each of their applications that they have to individually vet and curate and control.

Chris Bryce: Yeah, and I think it's a great topic within the CMS community, regarding what the integration of AI really means. And I think you described it really well, where there's some tooling where it's basically some chat within the interface. But I think that is a mandate of larger organizations, to have their own LLMs available to their different service tools, of which one is a CMS.

Chris Bryce: How are you enabling AI within the authoring experience itself? Are there different options for tooling? Do you bring your own? Maybe talk a little bit about that.

William Borg Barthet: Yeah, absolutely. So AI, the use cases are actually, every week that goes by we're finding new use cases for it.

William Borg Barthet: The first use case is, of course, that editor productivity, the chat interface for translating content, generating content, checking content, is one thing, but we're also seeing other use cases crop up. And I'm not talking about just images, but even things like content lifecycle.

William Borg Barthet: So if you're going to ask, hey, is there any other content like this one in the CMS already, or what content hasn't been changed for the last five years? Or even in the more visual end of it, where you're creating the visual end of the components, we're not there yet, but we're seeing also some CMSs generate the HTML for the actual visual parts of the rendered experience via AI.

William Borg Barthet: Now this can go in many directions from here as well. Perhaps what we'll see is, rather than AI being used at editing time to speed up the productivity of the editors, we might also see AI being used at render time, on request, to automatically generate super personalized responses to each site transition and each site view, which is another interesting direction.

William Borg Barthet: But I haven't seen that being used anywhere in production just yet. But it's very exciting to see where these innovations take us.

Chris Bryce: Yeah, we were chatting with a bunch of CMS experts about the idea of the entire experience being rendered by AI. Exactly, we're actually not that far away.

Chris Bryce: If, let's say, you're using ChatGPT and that memory can be shared, the context of the memory, then we just need to get things rendered a bit quicker. Why not display a picture of me on a boat with a pair of those white Adidas I've been looking for? Yes, totally. Exactly.

William Borg Barthet: Yeah, and the reason we're not doing that yet is not technical. The reason we're not doing that yet is that, especially, large enterprises aren't ready to trust the AI sufficiently.

William Borg Barthet: And this is where I think we're still trying to find the legal and cultural expectations of AI responses. Right now the feeling is that whatever the AI generates on behalf of an organization is what the organization is saying, and if the AI says something that's going to get the organization in trouble, the organization is in trouble.

William Borg Barthet: But I expect that as AI becomes more and more widespread and ingrained in our experiences, we're going to tolerate a bit more errors, and perhaps through legal disclaimers we can get, on request, an AI generated response that enterprises are happy with.

Chris Bryce: Yeah, well we had an interesting situation where we were building some voice agents for a large shopping center. One of the worries was that people would be uncomfortable talking to a robot, but the voices are so nuanced, and our solution was to give them a bit of humor, and have them start off by saying, I'm a bot, have you ever chatted with a robot before? It's really cool.

Chris Bryce: And people are excited, and what we found out is they're asking, they're trying to trick it and have some fun with that stuff.

William Borg Barthet: It's good, it's come such a long way. If you can remember the automated voice menus on a company answering machine, this is completely different.

William Borg Barthet: The language is completely natural. It's very dynamic, and you don't need to talk back to it like a robot so it understands which option you mean. It's a completely different experience, and I think the reason people don't like talking to robots is that prior experience, but not this new experience, which is much more enjoyable, as you say.

Chris Bryce: It is fun. I've got a bit of a use case maybe I could share with you. We have a manufacturing customer, and one of their challenges in a large migration project is that they've got a structured data challenge at the schema level. So they've got a bunch of product information coming from a PIM system.

Chris Bryce: They've got some attributes to the product in content, and they've got some other stuff coming from commerce. And one of their questions was, how can we use AI within the CMS, because that's what we were talking about, to be able to aggregate and synthesize schema level data on the page? That's essentially going to be kind of a static HTML page in a headless architecture.

Chris Bryce: What would you guys do with that? That's quite a...

William Borg Barthet: Challenge, absolutely. Yeah. So, in fact, put AI aside for a moment. In general, when you have integration with product information systems, there's two main strategies, a good one and a bad one, in my opinion.

William Borg Barthet: The sort of anti-pattern, architecturally, would be to replicate the product catalog into the CMS. You do this by batch job, and it's quite expensive in terms of resources. You end up with race conditions, you end up with issues if the downstream content is modified, do you override downstream content or do you respect this stuff coming in?

William Borg Barthet: So one strategy would be replicating the content in the CMS, so then you could decorate it with additional CMS content. The other strategy is a referential strategy, where you use external content references to reference CMS content to that external PIM data, and then fetch it on request and add additional CMS data to that response.

William Borg Barthet: Now, of course, there could be issues with performance, but with caching layers and other strategies you can make that perform. And this negates the necessity to replicate the product catalog in the CMS and makes it perform much better. Now, what does that have to do with what you're talking about, which is AI?

William Borg Barthet: I would suspect that a similar kind of architecture would work well with AI here, where, instead of dumping all the data ahead of time, you're fetching things sort of in a just-in-time fashion, and then checking to see if there's any references stored in the CMS or other touch points in the infrastructure, and then bringing it together in the response dynamically.

William Borg Barthet: I think that would be a better architectural pattern, most likely, but I don't think we're there yet with AI to do this, perhaps. There are some solutions out there, but I think that would be a very interesting custom solution and use case for AI.

Chris Bryce: A lot of CMO marketing folks are curious, and a bit uneasy about, people's adoption of using AI to find their websites. We see, like on LinkedIn, everyone knows how to solve this problem magically, and I have yet to find anyone who says, I have actually done this, and now we are found in AI search.

Chris Bryce: Do you want to talk a little bit about your perspective on that, and how that relates to the actual functionality or features within a content platform? I just know that's on everyone's minds.

William Borg Barthet: No, absolutely. AI is where search engines were 10, 15, 20 years ago. You've got search engine crawlers that scour the internet following links, finding content to index. AI LLMs are very much the same, right? They're looking for content and indexing it.

William Borg Barthet: Now, over the years, search engines developed best practices. We have things like robots.txt to guide the crawlers on how to index data, and we have specific metadata in the responses which guides the crawlers. In the AI space, we're very much at the beginning of that, and there really isn't any sort of best practices that have been established in terms of guiding these AI crawlers on how to find and process your HTML, essentially. That's in the public domain.

William Borg Barthet: But what these AI crawlers are doing is they're replicating a lot of the same things that the search engine crawlers are doing. So if you have good SEO hygiene, the right amount and types of metadata, and the right visibility from a machine point of view, like a sitemap.xml that's in good order, this is what the AI crawlers will be using to find your data and get it rendered.

William Borg Barthet: So if you're not getting found by the AI chatbot, chances are you're not getting found very well by search either. And it could also be that the crawlers of these AI chatbots are quite manually curated at this point, and they're heavily weighted towards Wikipedia, Reddit, and specific sources that the AI LLM curators have decided are more important than Google search results.

Chris Bryce: Right, you had mentioned in some of our offline chats a discussion just around the MACH Alliance. Oh yeah, and you want to talk a little bit about MACH.

Chris Bryce: I'm going to make an assumption that a fair amount of our audience isn't quite familiar with MACH yet, or they're just getting the hang of what all that means. Do you want to explain that a bit more?

William Borg Barthet: Sure, absolutely. In the last five years, maybe a bit more, there's been a massive architectural fashion of moving towards microservices architecture and headless front ends.

William Borg Barthet: In the CMS world, this fashion has resulted in many companies wanting to go headless, which means creating a separate frontend application rather than having their content management system render the webpages directly. It also moved a lot of companies to SaaS CMSs rather than PaaS or on-premise CMSs.

William Borg Barthet: This has been fine in many cases, but we're seeing perhaps a topping of that trend, and even maybe signs of a retrenchment. We're seeing some large enterprises, and particularly more traditional enterprises like universities, police departments, governments, questioning that architecture.

William Borg Barthet: And wanting to go back to a model where they have more control over their application, and also questioning whether the investment of having a separate application to render the front end is really worth the investment, because what they didn't perhaps anticipate is the amount of the total cost of ownership of having an entire other application in your infrastructure that needs to be maintained, upgraded, and hosted.

William Borg Barthet: Right now it's quite anecdotal, but we have noticed the trend. We've seen many CMS vendors create a SaaS version of the CMS. We've done the same, and what we're seeing is a renewed interest in our past on-premise version, and steady interest in our SaaS offering.

William Borg Barthet: Anecdotally, I've heard that other vendors are seeing similar trends starting to emerge, even though headless microservices architecture remains the dominant fashion.

Chris Bryce: Yeah, thanks for that. Personalization always comes up for us in a consultation of a migration or a modernization. It's one of my most interesting conversational themes.

Chris Bryce: After we get through kind of, wow, here's all the stuff we can do, it's who's going to do it? You need now a personalization team, you've got to resource that, create variants, all this kind of stuff.

Chris Bryce: Maybe do you want to talk a little bit about Bloomreach's response to personalization, and what does it look like to actually activate that, whether it's speaking to a DXP or a data platform, or how all that works...

William Borg Barthet: Yeah, absolutely. So in our CMS product we have personalization, but what it really is, is an integration layer to a personalization engine, a customer data platform, or some other system that segments your user base. The truth is that if you're going to do this at scale, and with large enterprise it's always at scale, you're going to need some kind of AI driven machine learning approach to creating the segments dynamically, or making it very easy for your marketing users to create segments based on the data coming in.

William Borg Barthet: Otherwise you're really sort of trying to pull this information from the ether and figure out what segments you actually need. So what the CMS then does is integrate very seamlessly with a platform like Bloomreach's Engagement platform, or Adobe Target, or whatever other personalization engine there is.

William Borg Barthet: It puts in the segments dynamically and then allows the CMS editors to target those segments with different content. If you want to scale that even further, what some marketing automation platforms are doing is doing even the targeting automatically, trying to figure out which content to show without that further interaction, and automatically testing those interactions.

William Borg Barthet: That's quite difficult to do server side, whereas many enterprises would prefer to curate that experience manually in the CMS, and by doing that have a server side personalized response, rather than relying on DOM manipulation in the front end to switch things up based on what the marketing automation tool thinks needs to be shown after the render.

Chris Bryce: You mentioned CDP, or a kind of data platform. Yes, Bloomreach is interesting in that, from what I understand, it kind of synthesizes commerce, content, data, and personalization.

Chris Bryce: Do you want to talk a little bit about how you differentiate against the couple of products you also mentioned, Magnolia and Adobe? Roughly similar engine underneath, in the core, in a way, but the rest of the finish is quite a bit different. I'm not sure if you want to talk about that a...

William Borg Barthet: No, absolutely, Chris. There's a lot of the same, you know, arrows in our quiver, as you mentioned. Our customer data platform, called Engagement, and our search and merchandising platform called Discovery.

William Borg Barthet: They're very synergistic with our CMS platform, and they use AI and machine learning to that end. The main difference, I think, between us and something like Adobe, is Adobe is still very much a suite solution, whereas, although we have multiple products that work very well together, we still see ourselves as a best of breed approach, where we're very happy for you to use all of our products, but we're just as happy to play nicely with others.

William Borg Barthet: And I think that's the main difference between us and Adobe, that also you will need other tools in your arsenal. We're not going to be your e-commerce platform, for example.

Chris Bryce: Are you open to sharing any case studies? No need to name names, but some success stories that maybe have a little bit of a unique storyline to them?

William Borg Barthet: Yeah, absolutely. I should have maybe done this before the podcast and gotten some clearance from some of our very happy customers to share their exact names, but unfortunately I've not done this, so I won't be able to name names.

William Borg Barthet: But we have some excellent case studies with large global banks. There's excellent work with some national police departments that have really personalized their web experience to the neighborhood of where you live, to show things like crime statistics. We've also got very large scale organizations, like national health services of major countries, that really try to bring together hundreds of websites into the same backend editing experience.

Chris Bryce: Cool. I'm just wondering, how would you define the mindset of the ideal customer for Bloomreach? What's going on in their world, what are they thinking, and where do they start to make a connection to considering Bloomreach?

William Borg Barthet: Yeah, absolutely. So these are always large enterprises.

William Borg Barthet: They usually have a problem with their web estate. They've got hundreds of websites, sometimes they don't even know all the websites they have. They're trying to pull them all together. They've got teams going rogue, creating stuff on more user friendly platforms like WordPress or Wix.

William Borg Barthet: And they're trying to create brand consistency across all of these different estates, and do this in a cost effective way, at scale. They're also very concerned with time to market, they want to make sure they can do this very quickly, and increasingly, especially in Europe, they're very concerned about server custody.

William Borg Barthet: All of the large cloud hosting platforms are US based, and I have no issue with that myself, personally, but I know a lot of large European enterprises are starting to get nervous and they're looking for alternatives when it comes to hosting their web, their CMS platforms, and such alternatives are pretty few and far between, although I know there's many investments being made in this regard as well.

William Borg Barthet: So the option of server custody, or to host on any cloud of their choosing, is also increasingly becoming a concern for our large enterprises, particularly outside of the...

Chris Bryce: US. Okay, interesting.

Chris Bryce: What is, I guess, the onboarding process? So if a curious opportunity is starting to consider Bloomreach, what happens next? What's that journey like, to onboard?

William Borg Barthet: Absolutely. Well, usually there is an RFP or some other process. You can reach out to us directly and we'll get in touch, give you a demonstration of our capabilities.

William Borg Barthet: We can run a proof of concept, but generally there'll be several other vendors in the race when it comes to these large enterprises. And when selected, train, onboard, often an implementation partner is also involved, and then we're off to the races.

William Borg Barthet: But if anyone is interested in our CMS capabilities, you could go to bloomreach.com or reach out to myself directly. We are very happy to put you in touch with our sales representatives, organize a demonstration, and answer all your questions, to make sure that this is the right platform for your specific needs.

Chris Bryce: Super cool, and we'll put all of your details in the comments below. We're just about out of time, actually, but I got a bunch of stuff I wanted to ask you. Maybe in the remaining time we have, are you free to talk a little bit about your roadmap? Where's Bloomreach going? What's some kind of the philosophical...

William Borg Barthet: Absolutely. In terms of our other products, I'll keep that light since this is a CMS podcast. In the CMS, our roadmap is dominated by AI. We already have AI in the product, we have an agent interface, but there's so many other use cases for AI that we're investing in, particularly, as I mentioned, in the content lifecycle space.

William Borg Barthet: And also in analysis at the web level, where you can look at the whole rendered DOM, rather than the documents behind the rendered page, and say, hey, does this webpage look like it conforms to AAA accessibility standards? Does this match our design system? So just being able to feed the resulting DOM to an AI, an LLM that is aware of the context of your organization, is what we're working on most rigorously at the moment.

Chris Bryce: Okay, so there's no worry that the web is dead yet?

William Borg Barthet: Not yet, no. But that is an interesting comment, because it could well be that the interface of the future is just the chatbot, and that changes everything. But perhaps you will still need the system that organizes your content and makes it available to the AI agents out there to get that data.

William Borg Barthet: And you need the mechanism to turn that off when you need to, make adjustments, and perhaps notify the agents when they need to crawl your data again. It'll be very interesting to see what system would emerge as that content management system for an AI world.

William Borg Barthet: But I suspect it wouldn't be any of the existing CMSs you see today. I think the architecture would have to be substantially different.

Chris Bryce: Right, interesting things to be thinking about, but in the meantime there are websites to be built and grown and extended and created. We are just a little bit over time.

Chris Bryce: William, thanks a ton for making some time out in your evening for joining us here, I really appreciate it. And for everyone who made some time to listen in, either now or in the future, if you do have any comments or questions for William, leave them in the comments of whatever platform this winds up finding you on.

Chris Bryce: And I'll just wish everyone a happy lunch hour over here in North America. Thanks again, I really appreciate it, and had a great chat with you. Super cool.

William Borg Barthet: Fantastic, Chris, thank you very much. It'll be my pleasure.

Chris Bryce: All right, okay, well I'm going to click some buttons and sign off here. Have a great day. Cheers.

William Borg Barthet: You too, Chris.