Talking Drupal #527 - AI in Drupal

Nic: This is talking Drupal, a weekly chat about web design and development from a group of people is one thing in common. We love Drupal. This is episode 527, AI and Drupal. On today's show, we'll be talking about ai, new Drupal features and the future of AI in Drupal with our guest Jamie Abrahams. We'll also cover orchestration as a module of the week.

Welcome to Talking Drupal. Our guest today is Jamie Abrams. Jamie founded a Drupal agency that focused on native Drupal CRM and business applications. He got obsessed with AI and is now a co maintainer of the Drupal AI module and part of the Drupal AI initiative.

Jamie, welcome to the show and thank you for joining us again. Thank you for having me. Also joining us for the next four weeks, Maya Schaeffer community lead at Evolving Web and lead organizer of Evolve Digital, formerly Evolve. Drupal Maya is the community lead at Evolving Web, a full service digital agency, designing and building better digital experiences for their clients.

She's the driving force behind Evolve Digital Evolved, formerly evolved Drupal, a somewhat series that redefined open source community engagement, bringing together over 1500 participants across North America over the last two and a half years. Maya also serves as a board member for the Drupal Association, where she supports global marketing and outreach initiatives for the next two years.

Welcome to the show and thank you for joining us.

Maya: Thank you, Nic.

Nic: And joining us as usual, John Picozzi, solution architect at eam.

John: I mean, not as usual. I feel like I've missed over the last five episodes. I've missed one or two. So, happy to be here. Happy to talk about AI and always happy to talk about Drupal.

Awesome.

Nic: And now to talk about our module of the week, let's turn it over to Martin Anderson includes a principal solutions engineer at aa and a maintainer of a number of Drupal modules and recipes of his own. Martin, what do you have first this week?

Martin: Thanks, Nic. Have you ever wanted to expose Drupal's capabilities to external automation platforms?

There's a module for that. It's called Orchestration, and it was created in August of 2025 by Jurgen of Lake Drops in collaboration with Drees, who some of our listeners may be familiar with. It has a 1.0 0.0 version available, which supports Drupal 11.2 or newer. It is actively maintained, has security coverage and a documentation site.

It has 11 open issues, none of which are bugs, which is pretty good even though the site is only officially in use by three sites according to drupal.org. Now with the orchestration module installed, external systems can trigger Drupal workflows, call AI agents, and execute business logic through a unified API.

The module functions as a bi-directional bridge, so Drupal events like content updates, user registrations, or form submissions, can also trigger external processing using the orchestration module with the active pieces. Automation platform in particular was featured at about the one hour mark in the most recent DRE note from Drupal Con Vienna, and we'll include a link to watch that in the show notes.

The complex example DRE shows is pulling content from a WordPress site using AI to evaluate whether or not each post met certain criteria, and then conditionally. And then conditionally calling one of a couple of ECA functions. In addition to using AI to rewrite the incoming content to change WordPress terminology into Drupal iss.

Under the Hood, orchestration provides an endpoint that will return A-J-S-O-N list of services, including the properties that are needed for each service. The external service also needs to provide the username and password for a Drupal account so you can control what services will be available based on permissions for the Drupal user that will be used already.

Orchestration works with ECA AI Agents Tool, API, and AI function calls. There's also work underway for integrations using webhooks for integration platforms that aren't ready to directly support Drupal's orchestration services. Now, in his presentation, DRES mentioned that they are looking for feedback.

Specifically. They would like feedback on what platforms should have integrations available. So let's talk about orchestration.

John: Oh my God, Martin, I'm so excited about this topic. Not even kidding. So, in last week's episode, I don't know if you listened Martin, but we talked about well we talked about orchestration a little bit but we also talked about the service that DRIs had mentioned in the keynote, which the name is failing me right now.

Active pieces, maybe active pieces there. It was and this looked really, really cool as far as a, a pathway into Drupal to be able to kind of kick off processes and, and vice versa, a pathway out of Drupal to, to other services. As somebody who's used Zapier in the past, I was like super excited to see this coming to Drupal because I was like, oh my gosh, I can connect all the things and it's easy.

And even me as like not a developer or a, a developer who could be dangerous, could do this and like just kind of move things into place and have things happen. So, I'm actually, I actually suggested to Nic and Steven maybe we we do some orchestration on the talking Drupal website to kind of, you know, Drupal codify some of our processes.

But all in all very excited to see this, this come into Drupal and excited to try to find that first use case where we can use it.

Nic: Yeah, i'm planning on evaluating orchestration. Maybe when it gets a little further along. I mean, I know Jurgen's been putting a lot of work into it, so I don't think it'll take too long to get further along, but I'm definitely gonna be installing it and kind of evaluating it. 'cause it, it feels like between all these tools, we're starting to get back to the ability to build complex Drupal sites, point and click.

Right? I mean, there's, I I imagine there's going to be some rough edges for getting it to that point still. 'cause

John: you know, I mean, I, I think, I wanna be clear there, we've always been able to build complex Drupal sites. There's no shortage of complex Drupal sites. I think, I think we're just a big

Nic: click

John: well, right.

I think we're making it a lot easier and accessible to people that aren't developers specifically, or, or wanna be involved in the code to be able to do, do these tasks. Martin, I do have a question about orchestration. Does it, I'm assuming it works with other services, like, you know, obviously Dre's demoed active pieces, which is open source, which like, great, we love Open Source, but if I wanted to use it with Zapier, would it work or some other service?

Martin: So the, yeah, so the work with Active Pieces has really been partly to provide them the code so that their interface will basically be able to pull in that list of available services and then in the UI sort of prop for what are the properties that you need to, to give it, to be able to sort of properly connect to Drupal for other services.

And this is, you know, to to your point, one of the things that Reen, Jurgen are hoping for feedback on, I think they've got a list started already of some of the, the services they would like to provide. Mm-hmm. A similar integration with, but it sounds like part of the. That process will be getting someone on the other side.

So for your example, Zapier to say, yes, we will add the Drupal integration and have it right, connect to the orchestration piece and maybe even as they've done with with active pieces, you know, write the necessary code so that it gets pulled in properly and, and all of those good things. Got it. So yeah, it, it may end up being a non-trivial.

You know, set of work for, for each of those, particularly because it does require some of the, on the other side who's willing to, to do the implementation there. But you know, obviously we've got some very smart people working on it. But the, the webhook piece does potentially provide a bit of a fallback.

So as long as any of those services, like a Zapier can just arbitrarily fire a web hook, there is potentially a way for that to be the way be before the sort of formal Drupal integration is ready. Yeah. To basically have it fire that web hook. Yeah.

John: If you're very intentional and you know what your, what you want your web hook to do, you can just fire it and it'll, it'll take it and do it.

Got it.

Maya: Yep.

John: That is very exciting. Very cool.

Nic: Well, thank you Martin, as always. Another great on-topic module of the week. If folks wanted to connect or suggest a module of the week, what's the best way for them to do that?

Martin: We're always happy to have lively discussions about candidates for module of the week in the Talking Drupal channel of Drupal Slack.

Or folks can reach out to me directly as mandclu on all of the Drupal and social platforms.

Nic: Awesome.

See you next week, Martin. See you then.

Before we dive into our primary topic, I have an exciting announcement an issue that I've been working on for the last six months to get, allow themes to have object jointed hooks just landed. So, those would, the audience that are excited about that. You can now with 11.3, you'll be able to convert all of your themes and I think speaking of Jurgen.

He was very excited about this. He's already converted gin to use them. So I think when it gets into core, it'll be it'll be modern and, and ready to go. So I think, I think what this means is you don't need dot theme files anymore either. So it's kind of, kind of exciting. You'll be able to organize your code a lot better.

John: Nic, Nic Laflin flipping the Drupal theming community table over one, one theme at a time.

Nic: All right, so moving on to our primary topic. Jamie, can you just refresh us what is the Drupal AI Initiative and what's it been up to?

Jaime: Yeah, so, well we started with the AI module. So last time I was here I was talking about the AI module, and I think I mentioned how we really wanted to make the AI module, you know, how you had like the commerce, guise of commerce, and I think a lot of dral things in the past have been a, a specific company being the, the main driver behind an area of Drupal in the country space.

And we felt AI was bigger, needed to be bigger than that. It, it touches on so many aspects of Drupal that from the beginning with the AI module, we had a number of different organizations involved and a number of different people involved. But in reality for a while, a lot of the actual work was done by Marcus.

And Freely Give was pushing things forward because we had the kind of like space as a company, you know, because I, because I, you know, co-found the company, I could put resources into it, I can make those decisions. Whereas if you are like an individual contributor is, it's a little bit difficult. If you are organization that you know funds, you.

It needs to pull you off things. It's quite hard to be focused. And so with the AI initiative, we decided to like really target organizations that wanna push AI forwards to like really like, commit and to commit to contributing. So it's like the, the heads of those organizations are saying, we're behind this.

They're releasing developers or marketers or a lot of different resources. And so the AI initiative was a number of different organizations coming together to push Drupal AI forwards. But it's the whole organization committing to it. And so everyone involved is committing both some money. And also individuals as ft sometimes full-time, A lot of the time people go full at halftime, it's quite difficult to give away some, like, you know, your top developers full-time to something that isn't billable.

But a lot of companies are doing like half of one of their developers or two developers halftime. And so the AI initiative is now under the triple association. And it's an official body to push dral forwards, triple AI forwards.

John: So can you explain to us your role, because Marcus is kind of the tech lead.

He's running, running the development and reviewing things, making sure they make sense to the greater mission. But what are, what, what would you say your role is with the project?

Jaime: My role's a little bit complicated. So what we're doing now is we're splitting things up into focusing on product and innovation.

So there's going to be people really focusing on delivering very specific things that can be used easily more polished. And then there's gonna be another area that is focusing on innovating, pushing things forwards, looking at nuts and bolts of things. And I'm, I'm more focusing on that. So officially David Lynch, who is another co-founder, freely Give, is actually taking on the role of being like a product.

Like a product owner of the in in innovation section, and we're still exactly figuring out who's gonna be like the product owner in the AI initiative for the product section. And then a what's the word statement for tender. Basically we're putting out a role where people can actually bid for it, which will be a paid for role to be like a full-time project manager for both the innovation and the product side of things.

So it's gonna be like a product owner who's in the initiative and there's gonna be like a paid for person to actually push things forwards. And I'm working with David focusing on the innovation thing. And so that's kind of my role is pushing the core of like the AI stuff forwards.

John: So that other, that other role's going out to like RFP or, or That's it.

Yeah. Yeah, yeah. Yeah. Okay. And. So in the past, I think, I feel like you've been kind of like a little bit of a project manager. I know that you put together a pretty great dashboard that kind of shows what everybody's working on and you know, people can, people can look at that and, you know, I've always, I've always thought of you as, as more of the, the innovation ideas, ideas guy.

Right. Do you guys have a project manager that's kind of like overseeing the, the work or is that gonna be kind of the responsibility of those product owners?

Jaime: Yeah well, so the product owners will be like David and someone else, and that's, they're gonna be not, not that many hours. It's gonna be kind of like watching, like, like representing the Drup AI initiative.

Yep. It's gonna be this RFP that's coming out that's going to be providing a, a full-time project manager who will then be doing that. So a lot of the tools I've built with Drupal Star Forge, have been like tracking issues across the AI initiative. Mm-hmm. Yeah. And they may or may not continue to use those.

We're moving to GitHub, so who knows what that will do. I still think we're gonna need some external tools. I think GitHub on its own won't quite be perfect for the, just, there's just, so there's something like 28 modules that we're tracking now in the AI initiative that is like really core to what we're doing.

So I dunno how much GitHub will just solve those. But then, yeah, these, these new roles, these two people who will come on will be like doing this full-time and then working with someone like me or another person in the air initiative which hasn't been decided yet. So to be product

John: question on something you just said.

You said that you guys are moving to GitHub, but we are GitLab. GitLab. GitLab, sorry.

Jaime: Oh, okay. I'm talking about the Drupal GitLab. GitLab. Got it. Okay. That's, that was gonna be my

John: question. I was like, wait a second. Why GitHub and not GitLab, but Okay. Just forget it. No problem. Keep going.

Jaime: Well, we actually are on GitHub as well.

It's, it's so complicated the, because recipes we haven't really figured out how to do recipes properly. So like, one thing that's happening with Drupal CMS is there's actually quite a lot of work going on on GitHub where you put all recipes in just one repo.

John: Yeah.

Jaime: And then figure out things later. So we actually have started doing that.

We might move to GitLab, but there has been a, an attempt to make a Drupal AI product, like a place with all the recipes that is more experimental. Mm-hmm. I used to call it Drupal AI playground. It's complicated. We still haven't figured out like, is Drupal CMS going to become the. Do we need something else other than Drupal cms that will be AI focused, and Drupal CMS will be more for people without ai.

I still think we're trying to figure out across drupal org, how do we release more productized things? So the with Christophe, he has been really pushing the product stuff and he's re released a few recipes recently. Like, something to help with SEO stuff to help with managing your LLM text.

There's a recipe about categorizing media and helping to search and find media and using AI to help you with that. So there's a lot of stuff that we've built into Drupal CMS that has been my innovation focus for the Drupal cms AI stuff that. Stoning works with Canvas is really focused on if you are like rebuilding your site on top of everything that is like the most cutting edge Drupal tools.

And so Christoph has been working on taking those innovations and turning them into recipes that you can put on your site like right away. We're trying to really focus on like how can we make an agency or Drupal site owner just get started with something useful very quickly. And that's something that's been lacking at at this stage.

It's been a lot of our stuff has been like the classic Drupal thing of these really cool tools and framework stuff. I mean, developers can very quickly do things and we're trying to really build more like something like a product, something that you can just get going with. Got it.

Maya: I have a question.

Jamie, you were mentioning going back to the project manager role. Maybe some of the listeners are actually interested in applying for that. Could you explain what needs to be done in order to receive the RFP once it's published? In order to apply.

Jaime: So I actually don't know off the top of my head. A lot of it happened at Drupal Vienna, so there was a lot of announcements about it at Vienna and I think there's a post on Drupal org.

Asking about it. We have said you have to be an AI maker. So it's only the people who you have to first apply to be one of the AI maker organizations, and then they are getting something. So I think there's a Slack channel for AI makers. There's an email that's gone out where the RFP so speaking to Lenny from the Drupal Association or Dominic, those have been the two who have been really spearheading that side of things.

So a lot of the AI makers who have come on to do with like, you know, signing up to be it signing the contracts, figuring out who the FT is, that has been Dominic and Lenny who've been pushing that forwards. And then once they're signed up either me or Christophe will then speak to those makers and then, and speak to the developers and start pushing them towards where they can actually get involved.

So it's been, yeah, there's just so much going on, which is why, like, you know, in the early days I was doing a lot of. So I was sort of a project manager. I was in charge of UX stuff, I was doing solutions architecture. I was doing talks. I was doing a lot of marketing. And it's grown. And it does mean that sometimes I'm finding it hard to exactly keep up with everything that's going on, which we're working on.

Paul Johnson is being in charge of the marketing of ai with Matthew Matthew Saunders, I think. And that's, I think has been amazing as well. There's a lot more on Drupal org that you can see and there will be a lot more. So yeah, Dominic and Lenny are really good people to talk to about how to get involved organizationally.

Maya: Awesome. Thanks. Yeah, so maybe you could share some of the most exciting AI powered features in Drupal, in the Drupal community right now.

Jaime: So I'm, I'm super, super excited by the Canvas ai stuff. So that was our real focus. It's a shame it's taken this long. I was really hoping, you know, away every single at Barcelona, we had, we, we started, we actually had some demos like a year ago of AI being to do a full migration.

Okay. Like literally point at page, look at the design, start migrating the content, building a theme, building the layout. It was all working with paragraphs. And so I was, we were, we had something a year ago that kind of worked, so I was always hoping this would would happen quicker. But it's been complicated.

Like Canvas itself has, has taken a while to get to a place where it's at our release. So we're, we're, we're really close to that. Now it's got our, it's got our release candidate DRAL, CMS 2.0 will be, I think January is the goal. And so we've been working alongside as Canvas has been developing the Canvas AI tool, have been working with it.

And it's a very. It's amazing. I absolutely loved working with Canvas, but it's very different to the, to what we're used to in Drupal, it being react, it being in the browser, it being much, the UX is really focused. It's like really, it's a good UX for that task. In Drupal, we tend to make UX is just fit in with everything else.

So the main goal is that your module uses the normal menus and things look similar as the field UI as opposed to actually really thinking about like, how, how should I do this task? Well. And so our AI agents that we built with that have been really great at just building a whole page. And it took a while.

Like there was a bunch of times we had really cool co proof of concepts, but they didn't, they didn't look good yet. It was like you'd ask it to do it and it was, it, it worked, but no one would want this 'cause it just wasn't cool yet. And I think with Canvas ai, we we're really, we're pretty close to it. So, and I think by by 2.0 we will be there. And I'm just really excited by that, I think. But once you've, once you can, once you can use AI to build a page, it's something opens the doors to so much other things you can use AI to optimize things for. So

John: let me ask a follow-up question there. In your example or, or in your, your early test case, right?

You were like, oh, my greatest site from here into, into Drupal, right? Was there a step in there where you could like ask the AI to like do other things? 'cause like I always think of like, people always use that use case of like, oh, I wanna migrate my site from this to this. Right? But then I always think that like, anytime you do a migration, like content architecture or content structure is garbage.

Like there's, there's, yeah, there's content that needs to be cleaned up. There's, there's all this stuff that needs to happen. It's never just like, take it from here, drop it there. Right. So, I was just curious in the early, early days, was it more like, Hey, go get the content and then like run this set of like refine it, refinements to it and then create the site?

Jaime: Yeah. I mean, exactly. So that was the process we went through. So we started with this like one-to-one match. Yeah. Put a lot of energy into it and then gradually realized this is a little bit useless 'cause no real life migrations are ever like this. Got

John: it. It's a really

Jaime: cool proof of concept, but really early on we're like, oh.

You know, no one was gonna pay, no one would pay any money for a pure one-to-one migration.

John: Yeah.

Jaime: That might not be true. We've got some things happening in, in the works right now of like WordPress, Drupal, cms mm-hmm. Where they're looking for something that is like one-to-one migration, but really quickly, we found it's gonna be quite hard.

We wanted a, the dream was like a, a non-developer tool. You could click a button and just migrate the whole thing and everything would just work. Yep. In practice what we're now aiming for is, can we make it so developers are like migrating sites 80% faster?

Nic: Mm-hmm. Because

Jaime: like, migration is so painful and so expensive.

Yep. If you get 80% faster, but you still need a senior developer, that's, that's still a huge win. And as we build these tools, we'll get closer. So, we were talking earlier about orchestration and ECA and at Vienna. I, I in my talk, I showed off something called Flow Drop. Which is this really, it's a focus on a ui.

It could work with ETA, but it's, it's focusing on the UI of orchestrating workflows in this like, really easy, nice. Clicky draggy way we can understand what's going on. And I, I think this is gonna be key for migration because like when you think about migration, a lot of it is like you have a data source, like a page, and you can see a flow of I need to take the page and split up into bits.

I need to use AI to take this blob of text, highlight the bits I need to pull out. Some of it needs to go into fields. Mm-hmm. What we found really early on with migration is that everyone's using paragraphs and layout. Layout builder layout paragraphs. And a lot of the time you have a blob of text, but figuring out like, what do I put into what paragraph?

Like this image. Do I wanna put the image inside a paragraph in using Seek Editor? Do I want its own thing for an image? What is the layout I want to use? One thing about the flexibility of paragraphs is there's just so many different ways you could do the same thing. And it was so hard to make a generic way that we could build AI to create a paragraphs thing.

I, I think we really early on could get it so that you could, as a developer for your site and the way you use paragraphs and your thing do something. University of Edinburgh really did a lot of experimentation with that, but with Canvas ai, because we have a lot more structure, a lot more metadata, a lot more stuff that we can explain to AI why to use something one way or not that really helps with ai, like laying things out and.

Really early on we're gonna start doing things where we're gonna have like, we have something called the context control center. DRIs talked about where like we tell AI this is how this site works. So when it's migrating something in, it can do a check with brand guidelines. So it's not just doing a one-to-one, but it might, you know, do a quick check of this is how we do it, or this is a new way we like to do our layout.

So even though that one was that way, this is our new way because we're focusing on accessibility or responsiveness, et cetera. And so that will be our real focus towards Chicago.

John: Makes

Nic: sense. And does it provide some sort of like integrated auditing service? 'cause I mean, I find aside from the developer time, one of the big, big time sinks is figuring out how to build reports that content editors can easily use to like, make sure that things are moving correctly in the right place and look correct and like.

Need needing to be able to say like, Hey, these are the 50 pages that we made accessibility changes for. These are the 10 that we made brand guideline updates for, like is that being considered or is that,

Jaime: yeah. Yeah. It's definitely being considered. In the, like, innovation space. We worked on those things.

Like we worked with like the workflow module in Drupal. We've worked with like revisions to do it. We created a system that actually integrated with Git hub, I think it was at GitLab, where it like created it, it created issues inside GitHub and would allow people like to check what AI's done.

It would create visual regression testing. So we've worked in the innovation world on those things. And then the big thing we haven't decided on yet, so Dre released a really great blog post a while back called Rethinking Drupal CMS Drupal in a world of ai. He had these big vision, like this big vision of how Drupal itself could change.

And we are sort of November starting to explore all of those things because we're gonna start wanting, we're gonna really wanna push, like we, we know we have revisions, we know we have moderation, we know we have workflow stuff, but everything in Drupal, 'cause it's, it's so cus easy to customize and we are usually expecting like only one or two people, or if it's a lot, it'll be built like a custom flow for that one organization.

If we have like a whole bunch of AI agents making changes all over the place, we really need to rethink in Drupal, how do we do that? You know, if you, if you have a team of a hundred people on your Drupal site going around reviewing all your content, or you have a hundred AI agents, to some degree the UI is the same and we need to like make that good in the core of Drupal, it's not really an AI problem, it's like a Drupal problem to some degree, I think.

Nic: Yeah.

Jaime: And I think that's gonna become, become a focus and it's something, that's why I'm really happy that the AI initiative is a lot bigger than me and bigger than free to give and bigger than AI stuff. Because really to do Drupal ai well, we need a lot of people who know Drupal and content management and configuration management really well.

And some of these problems are so hard that I'm finding, getting my head around it has been quite complicated of like, okay, how are we gonna do this? Like we got Canvas ai, how are we gonna do lots of revisions? What if like multiple people working on something at the same time? What about conflicts?

What about diffs? What do, how do you do the concept of a diff with something that's so visual at Canvas? There's all this that we need to like really figure out between now and hopefully Chicago. So I think that's gonna be a big push.

Nic: So let's, let's take a step back for a second and I don't, I think it's hard to have been working in Drupal in the last couple years, or anywhere in the world to not hear about ai.

If somebody wants to kind of get started with ai, it, it's so broad. What's kind of a good use case or that you can kind of start to safely test AI on your site? Like, if you're managing kind of a medium sized site and you wanna start, your boss has told you integrated ai, where, where should they start?

Jaime: Cool. So I would say the simplest is if you store the AI module, you'll have some, you'll have the seek editor plugins. So that's like the really basics starting point where you can, in seek editor use AI to suggest content, summarize it, change the tone, translate things. This is something we released like over a year and a half ago.

So that was like, that was like the good starting point. Mostly you are saving time from copying and pasting from chat GBT. It's not doing anything fundamental to Drupal, but it's a good starting point. It's, it's relatively safe. Stage two is Tri Drupal CMS and the AI agents there, 'cause those are really built to like, work nicely.

They're, they're safe. We put a lot of energy in the whole flow of what they can do especially when Drupal CMS 2.0 comes out. The Canvas ai agents I think are really cool. You can do a lot with them. What we're hoping to do with this product side of things is take a lot of the things that we know are actually the really good places, but you still need a developer to get going and then, and make it so that non-developers can just get going with it.

So AI search is phenomenal now. I think it's really got good. It's not yet stable enough, like the AI search module itself is still an experimental, it's not stable enough that we can I could recommend a non-developer to just start with it, but we have seen with a relatively small amount of developer work ginormous benefits with very, very low downsides.

So one thing we made was a AI. Views, filters, assistant. So it's, it's like AI search. You know, you've got a box, I wanna search for something. And we were looking at some medical papers and we were looking for, can you give me all the studies about lung cancer, not in Europe, for European Commission website.

And every single tool that it currently exists failed. So their existing site, if you type that in, it would immediately show you a European thing, but also Google didn't do it. So if you do Google site search, the first thing it would show you was something in Spain. Chat, GBT web search couldn't do it.

Because it was doing a normal search, looking at top five results, it failed. So we built this little module that just literally took your, the, the prompt that you said the question, looked at the filters that were there, and just filled in those filters. I mean, you just passed it into the URL and then you just saw a search with your sale term term, but all the filters applied and immediately it got it, got it right.

And there's no hallucinations. There's no privacy issues. It's just your content. And I think things like that are gonna be really powerful as a place that we can get started. AI search was a bit unreliable a year ago, and it's really good now if you implement it correctly. Hmm. So those are the

John: three areas, is that AI search fully in Drupal?

How do you, your database, like use using a, like using Drupal search and then layering AI on top of it, or is there some other thing going on in the background there?

Jaime: Well, so that example I gave the v filters ones, there's, you are using your, it's fully normal Drupal. There's no vector database, there's no anything else.

'cause all it's doing, the AI agent is just taking your query and just filling in the filters. And by filling in it puts it in the URL Like you can do with, with views, you know, you can take any, you can do all the filters, copy and paste URL and give it to someone else. Well, AI's basically doing that for you.

So there you do need an, an AI LLM provider. But apart from that, you actually can just get it working on your current, existing site. Mm-hmm. And you don't need anything else. You don't need any more indexing. You don't need anything. Yeah. There's other ways of doing AI search where you have to have a vector database.

There's other ways of doing agentic search, which I, I showed where you actually have agents do the search, look at the results, and then iterate over and over again, like trying to do searches. Yeah. So there's some more advanced ones that are coming. But all of this space of AI search, I think it, it, it, it can make things immediately better for end users in an area that's been quite painful for a large Drupal sites with relatively low downsides apart from, we're not quite there yet, that it's like a one click in store.

It's, it needs some developer. I think it's, I think it's funny

John: that it took AI to fix Drupal search. Just, I'm just leaving that there. We can move on. But

Jaime: I mean, it's been like that with the UI as well. Like the Drupal UI has been so complicated for ages. And then that was our first start in Dr. Drupal CFS 1.0 AI's really helping get, get, get some of the f some things Drupal's really good at and some things not so good at.

And one thing that's been so cool about AI is like DRIs kept saying in, here's Dre note that Drupal's sort become accidentally the perfect place for AI to happen because we were working so much on certain things. Were really good. But everyone would hate the initial experience. Everyone would find the initial experience of the, for the ux, tough.

They liked Drupal the more they used it. But the beginning was difficult. Yeah. And what we found is AI's been really good at solving some of the things that we've struggled with in Drupal for ages. Whereas when you look at something like WordPress or Squarespace, they've been doing one thing really well that AI can come into.

But actually it's not that easy for AI to work with, with some of those tools. 'cause it start, you know, once you get past the initial experience in WordPress, you are mostly just coding widgets or buying them. Yeah. So now you are moving into pure, using AI to write your code for you. In which case, why would you use WordPress?

You could just use, just create a hole. Vibe coded thing from scratch. Yeah, so I think, yeah, like with AI search and with some other things, AI is solving a lot, configuration management would be really nice for AI to just solve. And I think that one, we're gonna actually have to write code and do some proper architecture for, unfortunately.

Nic: Yeah,

Jaime: probably.

Maya: I think this all sounds really interesting. I'm really excited about the Drupal canvas as well and how it's gonna turn out and make things hopefully also easier for marketers and end users. You know, and like it, we can probably talk about like, how to put this further out. Like this will be my role at the Drupal Association as well to make sure that, you know, not just the developers hear about that, but also really organizations.

And so that's, that's one part. But I was, I was wondering as well, how. Is Drupal, ensuring that AI helps creators rather than going rogue. Like I, I know that Dre has talked about the main te of Drupal AI being keeping a human in the loop, but you know, you never know.

Jaime: Yeah, well at the moment actually with Canvas ai, it, 'cause it worked in your browser, it's not, it, it actually has to have a human in loop.

It actually can't work prop that easily in the background. And so we are actually working, we showed some aspects of background agents at Vienna and Chicago are working on them more. But at the moment, humans are very much in the loop. 'cause it's, it's, it's almost like, it's not quite this, but when you, when you ask something in Canvas AI to do something, you'll see AI will start thinking and then it will tell you, it's starting to build the page and you'll see it being built in front of you.

You see like the, the components fly in, it's, goes behind the scenes, searches your media library for relevant images, pulls them in. But it's doing everything in Canvas. And the way that Canvas AI works is like, it's got a really good like, click undo. Like it keeps track of every op operation you've done.

So it's keeping track of everything AI's done. So you can see all the steps, you can undo, you can roll back to a particular place. You can choose to publish the whole thing or not. So already with Canvas ai, it's very, very human, the loop. What we wanna do is actually go and try and get more automation so that AI can start just making changes to pages.

And then we're gonna try and explore how do we show that to a human. But it's the same thing of if you have a human working on a page and you have an edit an editor, how does the editor really know what that person's done? So that is a problem that we need to solve in Drupal anyway. Well, we are, we have been solving in Drupal, but I think we need to, to really do a much better job of it.

Much more holistic one. Yeah. Because if we're gonna trust these AI agents to actually build it, we really need to make sure those editors have good tools. Yeah. And so by solving that, which we wanna do in a content management system anyway, for teams of humans and editors, aI agents will, will do that.

We, we've actually made it so that it's not there yet, but by, by Drupal CMS 2.0, every single AI agent will have a role. We might even give them a user. That certainly they come up as a, as a, so you can see it's edited by AI agent, I dunno if we're gonna make it. So they all have their own, like personalized names and they're all separate users or if there's just AI agent, but they use Drupal roles so they all have the same permissions.

So they can't, the AI agents can't do anything unless they have permission to do that thing. And that also helps them fit into to workflows. 'cause you might have it so that like an AI agent can edit something, but they can't publish, for example. And then you want the Drupal normal flow to then allow you to see what they've done.

DRIs has been referring to as a human control center, his dream is someplace like a, like a dashboard that really shows you all these AI agents, what they're doing points you in the direction of like. Things that you need to check changes they've made how much it's been costing, et cetera. So we we're hoping between now and Chicago to push this human control center forwards which would help solve that.

Nic: Yeah, it's definitely, it's definitely a tricky problem because I, I think part of it also is you, you were comparing it to a human editor, but I think there's a couple key differences. You know, one is volume, right? It's not human. Yeah. Agent. Yes, that's one. But an ai, AI agent volume could be much, much higher than a human.

And also generally humans get better at a given task over time. Right. So you can kind of get an idea of like, this particular editor is good at this piece. I don't have to check that as closely. This piece they struggle with. I need to check that more closely. And yeah, the AI versions change over time.

But relatively, they stay above the same unless you're changing the prompt. And they also have the inverse problem of the versions are changing so quickly now, it's almost impossible to keep up with what that current skill is. And they might have to release a, a, a safety feature that reduces their quality again, dramatically.

And you have to, so having kind of an overview of like, this is what's currently happening. You kind of have to review everything always. And finding a way to make that simple and smooth will help the human editors as well when they're working with human editors as well. But I think it's crucial for, for this piece

Jaime: yes.

So I, the word I would think about that problem that you arise slightly differently. 'cause I, I would say the biggest issue we have in Drupal is how customizable it is out the box. So it's very hard for us to do anything nice in a product thing because like, we wanted to do sort of really simple MCP recipe for Drupal CMS.

So MCP is like a, is a special for AI way of telling AI agents how to interact with your sites. You know, what tools you can use. And we thought, let's make a simple MCP one 'cause WordPress released their MCP stuff and it had loads of press about it, but it was the simplest MCP thing in the world. It was just an MCP endpoint that allowed you to look at published content from your WordPress blog.

I, we, we did that a year ago. So let's put into Drupal CMS two. We support cp and immediately we came up with the problem, which is actually Drupal CMS one had some content types in it. Now it doesn't anymore or it doesn't really have any. So you create something, a really simple thing. It's like basically click a button to say, yes, this content type, I wanna allow AI agents to see published content for it.

And then even that was quite hard to dribble CMS because we had to like apply it to existing content types. And the same thing with why AI search has been so difficult is setting it all up is, is is not that hard if you know what you're doing. But every single Drupal site has a completely different set of content types with different fields that you would wanna index in slightly different ways.

So I've been working on AI agents that could like look at your content types, create those, you know, create indexes, figure out like how I index all those things, make views, et cetera. So you could have something where you look at the content types on your system, you install this thing and you have a thing out the box.

It's been very hard to do in Drupal. And I think the same thing with the problem that you've mentioned. If you have, like, you know, I've worked with some customer service agent teams and they have a constant flow of, you know, they have like 400 people doing these customer service stuff. Usually in the holidays they're like, they'll be there for six months and then they'll leave.

To some degree, the people, the QA people, and the managers have a similar problem to everything you described with AI agents. The people, they don't really learn because there's a constant flow of new people that you are, you're training and every single person has different idiosyncrasies. So when you look at that world, we actually do have a lot of good tools that exist.

And even in a Drupal space, you have, you have some Drupal sites, enterprise sites where there's ginormous teams of content editors Yeah. Making things where you have to look at them and there's a big churn. Yep. The problem is. It's very hard to get at those tools because every single enterprise Drupal site will use a subtle different way that they customize the workflow module or customize this.

And so the core Drupal experience doesn't scale very well. It doesn't really help you look after teams of a hundred content editors that you may or may not be able to trust. Generally the core Drupal experience is good for relatively small teams where everyone, you could give them basically admin permissions or sub admin permissions and you don't think about it.

So I think we really need to think about, it's more like a configuration, configuration management, having more opinionated stuff, and then building UIs like with Canvas AI that do this like moderation stuff really well and focus less on making it customizable for every single possible edge case and more about it actually being good for most of them.

Yeah.

Nic: As long as one of those edge cases is translation, I'll be, I think I'll be happy because that's one of the ones that ends up getting left off on those edge cases. But yeah, I, it's a hard problem, right? It customizability is kind of what's core to Drupal, but it makes these types of initiatives very, very difficult.

Yeah. I, I do wanna pivot for a second and I'm curious how the AI initiative or Drupal in general, you're seeing how are they're approaching transparency and ethics around AI generated content and decisions? 'cause there's, I think we talked about this last week, part of, part of the issue isn't just quality control, right?

It's also accountability, right? Who's accountable for the content going out into the site? If. It's just agents doing most of the work or a piece of the work. So how, how is kind of Drupal working on that aspect of this?

Jaime: I think it's, it's sort of the same problem to what, what, which, to what Ma Maya brought up.

'Cause it's the same kind of thing of. Of how do you, I I think it's, I think it's literally identical to how do you deal with, you've got 200 human content editors and they leave after six months and there's churn. We need to make those tools better. So we, we created like an early thing where we use Google our Lighthouse, and it looked, it had a background agents that looked at your page and we had it with some actual live enterprise clients.

And we looked through your page and we found we were looking for like SEO issues and they hadn't put in the meta description. And so AI changed that. But it acted like a user. So it made a draft version of that entity. So you could go into the moderation and you could see, oh look, this page AI agent did this.

We actually got ai a, the AI agent, to write like, almost like a commit message, a description of why it made that change. So you as a moderator could go and see that whole list of changes. See, oh look, AI made a change. This is why, this is what the change was. And they could approve it. Approve it. And we use core Drupal moderation tools for that.

I, I think those core Drupal moderation tools are the place where we should solve this problem to do of accountability 'cause. It's the same thing really, that if you have a team of low end editors with lots of churn, you actually wanna make the, the manager team the one who's really accountable because they need to be checking things.

So you, you push a lot of the accountability up. And you do that in an organization now anyway with AI agents. You really need to do that. So we do have those tools in Drupal and we did start using them. It's just they don't scale that nicely yet. They tend to work nicer for like a handful of changes with a smaller team and we need to improve those.

And that is a big, it's a dream of DRE to really focus on that. It's one thing he really wants to push. And it's something that we definitely gonna need help on because it's not my core competency. For example, like my thing is I still do that kind of moderation stuff, but more for like business workflow, CRME stuff, not content.

And so it's something that we're gonna need a lot of help. From the AI initiative from the Drupal community to like make that stuff better.

Nic: So I'm, I'm gonna pivot again 'cause this is something else we talked about LA maybe we should have invited you on last minute ad hoc to be on last week's episode because we're touching on a lot of things.

But there's a lot of interest in the AI world, especially in Drupal, right? There's, there's no, no denying that the initiative is, has more energy behind it than almost anything except for kind of the other starshot related initiatives. Right? One of the things that I'm fi that I think Drupal core in general, kind of desperately needs is an overhaul of the batching, queuing system, right?

Hmm. A clean way to like manage and update and rerun and process batched things. And that's something that the AI initiative. Is probably going to need two is are there plans for like the AI initiative to kind of like find pieces of core, like content moderation, batch queue, API, all that kind of stuff?

That could use modernization? Could use investment and invest in helping refurbish them or modernize them?

Jaime: Yeah. Yeah. Definitely. Just before I answer that directly, going back to the previous thing, one of the reasons why we haven't focused on the accountability of content is that we actually haven't focused on AI generating content.

That hasn't been our focus for a while. We focus on AI helping you build a, build the site. There's a new module called Tool I think it is, and it's the tool API module that Jurgen has been working on with with Michael Lander, and he's really been pushing that. So in version 1.0, Marcus made his really weird hacky approach to tools which is like the things AI can use to do stuff on your site.

And it was really hard for anyone to use in version 1.1 of the AI module and AI agents, we have a more. A proper version of tool tools that now a lot of people are building them, but it wasn't that secure. So we created something called the Content CER tool, and it was a tool that AI agents could use to create and edit all entities across the site.

It was so cool. We had really cool demos. The problem is the user is an entity, so you could easily prompt engineer it to just create an admin role. Now we had tools to constrain these. We had UIs to constrain them, but you had to know what you're doing by default. These tools could, would just allow AI to like edit the password of anyone.

And it was very scary. So we never released it. So this tool module is like the place where we are doing that. But now we're, we're, we're doing it secure by design. So we're thinking every tool has to have permissions, roles, et cetera. And now what that means is we can actually start creating content.

So we're gonna have some demos. Michael's been pushing them to show you being able to create content with AI more reliably and securely, which then pushes this moderation issue. But the reason why intersects with ECA is because in core we have this, we have something called actions for views, bulk operations, but actions are, they're a little bit hacky.

There's this big issue to make them better, that Tim Plunkett started 13 years ago that we wanna improve. And what we need basically for AI is we need a lot more structure on what the inputs are. So let's say you have an input, which is like entity. Well, do I want the entity machine name or do I want the entity id?

A lot of the time with a Drupal action, you just have to guess. Maybe there's documentation, but a lot of time you just try it out and does the page. The page, like the site white screen. So what we need is structured both inputs and outputs. Similarly, actions just do something. A lot of the time, they don't tell us that they've done something in a structure that makes sense.

There's no way in Drupal core actions that you can see that. So there's a core thing that we want, which is linked to this batch stuff which ECA has been solving, although it also needed more metadata. So, which is why we have tool where we wanna make actions in core just have better metadata on what, how you input things into them, which will make it easier for developers as well, or make it easier for everyone, is like, instead of having to go and read the Drupal docs, it's in your API and more structure on the outputs and how those work.

And so that is, I think, something we've made. But the dream of tool is it should go into Drupal itself, I think. And that will be a place where we start tackling that in. In terms of the, going back to your batch thing. Now the actions are the things that happen, ECA and these other things are stringing together, actions together, which is sort of like what batches can do.

That's another thing that we wanna work on. So like with Alexander Open Social, he's been trying to push Async forwards, which has been really important, like Drupal Async we've done stuff with fibers, which has been really, really important. But that's one of my dreams with this flow drops slash eca, is that we, we are thinking that actually, when you look at everything AI agents are doing, we wanna make a consistent way.

Running tasks, tasks and I'm really into doing everything in Drupal and there's disagreements, but like Dre has been working with this orchestration module and active pieces where active pieces has its own way of handling batches. And so one vision is get rid of Drupal batch API use active pieces instead and have some external tool tool to do all batching of any task.

And we use adjacent API in Drupal to expose what Dral can do and just go actually a stop batch, API tool. Everything is gonna be done by this active pieces or something. What we're trying to do with Flow Drop is create a consistent way of doing orchestrating workflows, which will then become a good Okay.

Consistent batch. API runner. And then we're also exploring with ECA and Jurgen. Is ECA gonna become the core place? Should we be doing something else? Should we be doing something in quadruple? But yeah, it was, it was pretty, I I was pretty shocked actually when I got into vibe coding. I keep asking Marcus, like, build something.

It's like, can we make this by a batch? He'd be like, that's gonna take me like another week or so. It's just like so hard to do the same task in a batch. I didn't realize how hard it was in Drupal until very recently. So yes, definitely a focus that's almost in my little world of innovation, this side of things of orchestrating AI agents.

A mixture of AI and non-AI tasks together is like my focus for the next few months.

Nic: Yeah. And, and, and my, my specific question here is, and again, I'm, I'm happy to talk chat more in an issue because we're, we're starting to run up against time, but my specific question, like the batch API is gnarly. It's gonna take a lot of effort to rewrite that piece and make, but it is crucial.

It, how do we. Redirect some of the excitement in the AI initiative to the less exciting pieces that need to happen to make it work in a sustainable manner for Drupal, the Drupal project long term, right? Because, you know, one option might be to move it offsite, but that makes it, you're dependent on an external service and we've had problems with upstream dependencies before, and then it's a big lift if, if, if, you know, active pieces goes away for some reason.

Right. So I, I, is there a way to kind of find the pieces like config actions and or not config actions, actions in general and batch API where it's like we really need this to be shored up for, to unlock ai. Maybe it's something that the project manager can do when they're on staff, help prioritize that kind of thing.

But I think it'd be interesting to identify those pieces and figure out how to incorporate them to the initiative.

Jaime: Yeah, like, it's, it's a complicated one. We, I, I, I feel like the easiest place is to start with this actions thing because actions in Drupal are so bad, so obviously bad. The improvements are so obviously good, and it would just make things better for everyone.

In all non people who don't like ai, it would make things better for them. It will involve a breaking change, though, because it means our actions, metadata actions won't. Can we get anything? Even something as simple as this actions thing into Drupal core as an AI initiative in what we're doing is going to be like a thing that I want to sort of, I, I feel like it's gonna be a good starting place.

The batch stuff is, is a little bit more complicated. What we're finding. Any AI initiative? Well, the way I like to work is I'm very, very much like running ahead. I, like Marcus has found it sometimes a little bit difficult and some other developers have found it difficult to work on me because I'm only interested in things that can be done in like a day, maximum a week.

If I say, let's build something, it's like, yeah, that's a good idea. It'll take about three months. I'm like, that's the same as infinity to me. So I tend to like move incredibly quickly. I'm really excited by what can we do that will make something look good and build something that will work by tomorrow or by lunchtime.

And so we move very fast. And even with ECA and Juergen, even though that's outside of core ECA is used by a lot of people. It's a lot more structured. There's a lot more legacy stuff that goes along with it. So even working with ECA, the way that we've been doing things in AI just been like moving incredibly fast.

I mean, I've been pushing for more breaking changes, markets pushes back. So in the AI module we do try and avoid breaking changes apart from major versions. I personally don't like that. I'd rather we did more breaking changes. And if you are using it, it might be a nightmare for you 'cause you're constantly having to change code around.

But we need to move fast in the AI world 'cause. Everything is moving fast. I've got pushback and now we've got more project managers and we are now, we've got more of a release cycle and management around that. But the same thing with the batch stuff. So what I'm hoping to do is start with like flow drop outside of even ECA, figuring out like how to do that well quickly.

And then the question is, do we, do we work VCA to do it better there or do we even go straight to core or do we go into Symphony? So I've been exploring, like Symphony has some really cool stuff to do with, like, they've got a workflow tool, they've got a Messenger, API, they've got various things. And a lot of me have felt like I, I like active pieces a bit, but if we're gonna do things outside of Drupal, doing things in Symphony feels natural.

And if there's these core things that Symphony already does well. That feels natural. We're actually gonna be at an event in Paris, beginning of December. And Fabian Porta is gonna be joining a, we we're gonna do a whole series of Drupal talks. It's not completely confirmed yet but Fabian's gonna be joining us and we've been talking to Chris, who's this symphony AI person.

We have this dream of eventually actually bring some of the AI module get Marcus to help with the Symphony AI and bring some symphony AI into the Drupal AI module. So we're not doing everything in Drupal. That doesn't need to be done. And I wonder if this batch stuff might be another candidate for working closely with Symphony.

I don't know yet, but I think the thing we're gonna do first is work fast, be really experimental, do this stuff with flow drop, make it feel good, work well, and then figure out like, how do we do this? Like properly?

Nic: Yeah.

Jaime: Et cetera.

Nic: I, I, I can understand that.

John: So one thing that came to my mind earlier when you were talking about bringing it back to migrating content last, last year I was at Adobe Summit and they showed exactly what you were talking about, some AI magic that they were like, look, you can, you can migrate your open source website to Adobe Experience Manager and here's how you do it.

You press a button and it walks you through migrating your site Now. I was like, huh, look at that. That's pretty interesting. And of course, we didn't have, we hadn't made the advances we have made in Drupal ai to, to kind of combat that. But, i'm wondering from a, from a community perspective, how is Drupal's open source model uniquely positioned to innovate with AI compared to those proprietary CMS platforms?

Jaime: Yeah, so there's a couple of reasons. A couple of things. Firstly, I, I think one thing we haven't seen the benefits of it. That much yet. Although we have some amazing case studies on drupal.org now showing people use AI for real stuff. But one of the things I think when you look at Drupal compared to like OpenAI, is OpenAI is stuck in this little ivory tower building AI stuff, but they're really far removed from their actual end users.

And Thropic have done some insane things, like the way they built MCP at the start was never gonna work. I thought they were just throwing it out there because like, it was like a proof of concept. But those ways they've built it, that clearly showed that people made it really have never tried out their own AI models in the real world.

Nic: Mm-hmm.

Jaime: We are closer to the real world. So I think one big part of Drupal is that we can make things practically useful quickly, like this, this, this views filter thing. You can just, you know, we built it on the ECA European commission's own distribution of Drupal. So like we, we built it in such a way that they'll be able to just roll it out fast and that thing of, like, that our community can make things.

But then the other power we have is that because of what, the way that we are very modular and the way that Drupal AI ecosystem works, all of those things can come back into our ecosystems and spread throughout it. So if you think of like WordPress, WordPress is open source, but a lot of it's add-ons aren't, well, they sort of are, but not really.

So every single benefit is, is locked behind some paywall of one specific vendor who, again, is just making all their money, making that product. They're not on the edge, they're not actually working with people. So I, I think this is gonna be something that we're gonna see unlock the potential of Drupal AI as more and more people get involved in building things for their clients.

In sharing it, we all benefit from that. We're seeing that a little bit already. You know, you know, obviously we're free to give a lot of, we've done a lot of commercial work with AI now and that's flowing back into the AI module. And more and more organizations are doing those kind of things. My hope is as this recipe world starts, as things really take off, as the initiative takes off, 'cause it's only been around for a few months, we're gonna see a lot more of the stuff that companies do with their own clients flowing back.

And you'll really see the benefits of specifically the Drupal. Open source architecture, because even with Symphony, it's a framework, but the, the actual applications people build, they're not interoperable. So like if you do something really cool in your symphony application, it's not obvious that the symphony community can benefit from that.

Whereas actually in Drupal, almost anything anyone builds, if the organization is willing, we actually physically can benefit from it. We can look at how they built it and use those same tools. My my, the, the question to me is really, I'm really loving the AI initiative and the energy behind it, especially the money and fte.

Is this gonna keep going? Like, because the, the downside of it is that we do sometimes have that problem where, like, to me. I know there's controversy around Acquia. They've done so much for Canvas ai. They've put so much money into like pushing Drupal itself forwards. A lot of Drupal agencies will tend to like fund just them a module that is useful for that client and then sort of like leave it hanging until they get another client to pay for improving that.

Whereas this thing of like Canvas where they're like treating it like a product that they're just making good we need to see more of that in Drupal. We're seeing that in the a initiative, can we make that sus financially sustainable? Can we make it so that people who put significant effort into like really pushing core bits of Drupal, like, like its moderation forwards, can make money on the fact that they've done that?

Is, is is to me the the gonna be the turning point. For, for, so when you.

Maya: When you talk about the future and like, let's assume everybody keeps you know, being part of this, pushing, investing in it, where do you see, or what do you think, what does an I native Drupal site look like in three to five years?

How will it look different from the size that we can build today?

Jaime: So I think I think you're gonna have AI everywhere. So like everything that you need to do in Drupal, there'll be AI assistance that can help you make those things. I think that ev a Drupal site is gonna be a series of background AI agents just doing loads of things.

I really believe at the moment we use these big proprietary models. We're seeing a lot of evidence. I haven't had the time to prove it yet. We're seeing a lot of evidence that these small open source models can be finely tuned to specific tasks and actually beat the big ones. And so I actually see the future as not you work with just open ai, but actually there's a whole, but like, like we submit modules, we're gonna have hundreds of little trained modules, mo models, AI models that are really good at building fields, really good at accessibility, really good at making views.

And so you would have a Drupal site that is a collection of AI models. Working with you in the background presenting to humans, sort of like what they need to do. It, it, it'd be like a, you know, an army of these drones flying around you, like picking things up for you and you orchestrating them the whole time.

And the other thing is that I think agent to agent marketing is gonna become really important. We're already seeing that. So in the cutting edge enterprise marketing people I speak to, they are excited by canvas ai. But the problem they have is that's aimed for a human to look at and they're already finding a lot of sales is happening where large organizations and B2B stuff, they're using AI to help you find those things.

So I think that's gonna become the future and it's gonna be really important to have something like Drupal in that world. And it's why the context control center, I think is gonna become more and more important because what you find with humans is you need as a marketeer to take your message and condense it down to something as simple and as quick and easy as possible.

If you're gonna be tagging and meta tags, you need a small number of very short tags. With AI agents, you actually go the other way. You need rich metadata. You need descriptions of how this module should work and all the different possible ways. You need ginormous f banks of FAQs that go into lots of very specific niche data about your site.

And so I think Drupal, the future of AI to AI thing will be Drupal storing loads and loads of content that isn't meant to be read by someone, but it's meant to help you understand more and more about your organization, your goals, your purpose this, this all your products, loads of information behind it to help ai like get more information out of you.

So I think this con context control center will become a key thing that you'll be managing in Drupal, not just a page, but like all of that information that you currently have in your organization that you use to make Drupal sites will go into Drupal and will use a content management to manage. So

John: I think it's something like that.

So just like, just like today, Drupal is a content hub, could be a content hub for your organization. Like in the future, Drupal's gonna be like a content plus AI hub for your organization where it's kind of managing your, your interactions with with, with your AI services. Mm-hmm. Yeah, I think so.

Jaime: Cool.

Nic: Well, well, Jamie, thank you for joining us. It's always a pleasure to have you on to talk about AI and the cutting edge in Drupal.

Jaime: Cool. Well, I obviously, as you could tell, I love talking about ai. I so very happy to do this again. I'm happy to talk about any of these subjects. Yeah, we are on hash hash AI on Drupal Slack.

So we can have regular meetups there.

John: Do you have questions or feedback?

Reach out to talking Drupal on the socials with the handle Talking Drupal or by email with [email protected]. You can connect with our hosts and other listeners on the Drupal Slack in the Talking Drupal channel.

Nic: And if you want to be a guest on Talking Drupal or a new show TD Cafe, click on the guest request button that I will add back before this episode releases

John: you.

Yes, you Dear Listener can promote your Drupal community event on Talking Drupal. Learn [email protected] slash td promo.

Nic: You can get the Talking Drupal newsletter to learn more about our guest host show news. Upcoming show is a much more center for the newsletter at talkingdrupal.com slash newsletter.

John: And thank you patrons for supporting talking Drupal. Your support is greatly appreciated. You can learn more about becoming a patron at talkingdrupal.com and choosing the Become a Patron button in the sidebar.

Nic: All right, Jamie, if our listeners wanna get in touch with you, how can they do that? And what are, where will you be in the next couple of months?

Jaime: Yeah, so, I'm on LinkedIn so you can reach out to me there. And on dribble slack, on hash ai, we've got hash AI dash contrib, there's hash ai dash initiative for those things. But hash AI is the main place. So I'm, I'm there all the time, so if you wanna reach out to any of us you know, I speak to and Slack and I'm, I'm at events a lot.

So the next one is in p Pune, I think near UBA in India. So me, Marcus, and Shi who did the flow drop and QD 42 who are organizing it. And they've been really instrumental in AI stuff. They've done a ton. We're all gonna be there and it's gonna, it's this, it's called the Oasis, but it's, we're trying to make like sort of an elite Drupal AI event of non, it's not like first time contributors, it's people who've already done stuff and we're gonna have a couple of days and start actually building things.

It's gonna be like a hardcore sprinting talking about those things. And so I'm really excited by that. Drupal Star Forge AI is a good website to see an overview of everything we're doing. And I think we're gonna be at API days in Paris. I, I referred to that there. Awesome. So, yeah, that's this year.

Nic: Awesome. And Maya, how about you?

Maya: Thanks so much for having me, first of all. Yeah, I am also best reachable through LinkedIn triple Slack as well. And if you, what I was gonna mention, I am, as Nic introduced me earlier, I am the lead organizer of Evolve Digital as well, formerly evolved Drupal. The next one is coming up in November, November 20th and 21st.

It's the first two day summit, so super exciting. We still have an evolved Drupal track and then tech and AI track. So we are going to cover Drupal Canvas and other AI and Drupal topics. We also have a training there. Mike Canello from Triple Easy is gonna give a, a dr. AI basics training, so I just invite all the listeners to check it out.

Evolve digital com and yeah, early bird registration is still open until November 5th, so another week. So yeah, hope to see everyone there.

Nic: Awesome. And John, how about.

John: You can find me personally at picozzi.com or on the social media and drupal.org at johnpicozzi, and you can find out more about EPAM at epam.com.

Nic: And you can find me at nicxvan, N-I-C-X-V-A-N, pretty much everywhere.

John: And if you've enjoyed listening, we've enjoyed talking. Have a good one, everyone.

Nic: See you next

John: week.

Maya: Thank you. Bye. Thank you. Bye