Enterprise WordPress 2025: AI Tools, CMS Strategies, and Roadmap Updates

Tom: Alrighty. Welcome everyone to this week’s episode of Scaling Enterprise WordPress and Open Source. I’m excited today to be joined by Brad, who you’ll recognize from last week’s show, and also returning host Karim, who we haven’t seen since prior to Christmas. So, great to have you back, Karim. Thank you. Before I hand over to Brad to talk about what we’re going to be discussing on today’s episode—if you haven’t checked out last week’s or, I think, a couple of weeks ago now—definitely do so. Brad and I did a director’s commentary on the State of Enterprise WordPress Report that recently came out. Tons of interesting data in that, which we added some hopefully interesting flavor to as we scrolled through it. So, check that out and give us feedback if you have any. Brad, over to you—what do we have on deck for today?

Karim: Yeah, we’ve got a fun show today. Happy to see Karim back—happy to have the gang all back together. So, today we’re talking about what really lies ahead for enterprise WordPress in 2025. There are a lot of trends out there, a lot of buzzwords, and many things I’m sure we’re all seeing within our own agencies, reading about, researching, or hearing about at events. I want to talk about some of the trends we’re seeing, hearing about, and incorporating within our agencies to bring to our clients. I’d love feedback from anyone listening too, because of course, these are just trends and ideas of what we’re seeing and expecting, but by no means an exhaustive list. That’s the beautiful thing about technology on the web—it’s always evolving and changing. What we think is going to happen will be interesting to recap at the end of the year, or maybe early in 2026, and see how we did based on some of these predictions. So, some pretty big ideas and concepts we want to dive into—Tom, you want to kick us off?

Tom: Yeah, I mean, I thought we could frame this discussion around three major categories. I think the first one we’d all probably throw on the board is AI—it’s barreling towards us at rapid speed. Most of us are probably playing around with it ourselves, incorporating it into our work, and having lots of conversations with clients about it. It kind of feels like this is the year that real, impactful advancements are going to happen that will directly influence our work. There’s a lot happening in the broader CMS space as well, with proprietary platforms pushing the evolution of headless, composable, hybrid, and universal CMS models. The whole industry is trying to reconfigure, and WordPress is a part of that, so I thought it’d be good to unpack that. Lastly, there’s the WordPress roadmap—a lot on deck for WordPress itself, including full-site editing, which is making a splash in the enterprise space. We’re seeing Phase 3 of Gutenberg with collaborative editing likely this year, and perhaps even Phase 4 kicking off with multilingual support—that’s worth discussing too. So, I figure we could categorize it into those buckets and start with AI as the major trend. Where to start with AI—it’s coming fast. A couple of things that are becoming more real for me, as the CEO of an agency working in enterprise WordPress, are the degree to which AI is already affecting software development. Human Made does a lot of software engineering, and we’re starting to see a clear difference in capability and efficiency between teams using AI code-assistance tools and those that aren’t. That’s really sticking out for me, and it feels like it’s going to become even more important this year—especially with the release of next-generation models like GPT-4.0, which are getting better than the majority of programmers out there. I’m curious—have either of you seen your teams incorporating those tools, and how do you think it will affect software engineering teams we’re all hiring or collaborating with?

Karim: You’re right—AI is everywhere, in everything—whether it should be or not. We’re seeing massive saturation, especially on the development side. The big question is: what does it mean to be a developer a few years from now when AI tools are so advanced that they can generate code or create full-blown web apps in minutes? One challenge now is sifting through the noise because AI has become such a buzzword—99% of what’s out there is garbage. It’s about understanding which opportunities and tools are genuinely useful. Sure, there are big players like Copilot, which many of us are using or experimenting with, but there are so many additional services that could be amazing—or not—and it’s about sifting through all that to figure out what actually makes sense. Some of it is standing by to see who emerges as the dominant player in each space. From my perspective as an agency owner building large-scale websites, I see AI as very much an assistant tool—not just for development, but for everything. I use it for contracts, writing emails, and drafting responses—it saves time and makes my communication clearer and more polished. But one concern is over-reliance—making sure we don’t lose the skills needed to understand if what AI produces is accurate. I’ve dabbled with AI to build WordPress plugins and API integrations—sometimes it produces great results, but other times it uses outdated methods. For example, building a settings page—does it use the proper WordPress Settings API, or does it try to insert data directly into the options table with raw database calls? Less-experienced developers might accept it as correct because it “works,” but it’s not the proper method. So, one of my big concerns—and something my leadership team focuses on—is ensuring we’re not relying on AI without verifying its accuracy. I think many agencies relying too heavily on AI without proper oversight will produce substandard products.

Brad: I completely agree with that. There’s been a rush to assume AI will revolutionize coding overnight, but it often doesn’t write efficient code or even solve the problem correctly. Some folks out there are experimenting with demos that, in practice, don’t work as well as the tried-and-true methods we’ve relied on for years. From a business perspective, our enterprise customers are reacting very differently to AI. Some have outright banned its use—concerned about the origins of training data and potential liability issues—so they’ve asked us not to use it. We’ve had to explain our usage of AI for QA, automation, and efficiency, but also assure them that human oversight remains critical. Internally, the speed of AI advancements has dramatically shifted how we work. We’ve always kept a few developers on the bench for R&D and unexpected needs, but now, much of their time is spent evaluating new AI models and tools. There’s something new to test almost daily—it’s added significant internal costs in terms of time and resources. I’m curious—how long can this industry keep up this pace before we collectively decide to step back from the daily grind of testing every new release and shift to a more strategic cadence?

Tom: That’s one of the fundamental challenges—it’s all moving so fast. If you don’t test the new breakthroughs as they happen, your assumptions can become outdated. For example, someone who tried ChatGPT back at version 3.5 and found it underwhelming might assume nothing has changed. But we’re in a fundamentally different place now, and in three or six months, the landscape could shift again. It reminds me of the early internet days. Back then, we were iterating on technology rapidly, but without the global spotlight. Now, everyone’s watching—AI is helping us draft emails, respond to queries, and ideate. Can we keep up with this relentless pace?

Continuing from where I left off, here is the completed transcript with corrected spelling, punctuation, longer paragraphs, and speaker attributions:

Brad: As our resident historian, I’ll say the big difference between now and the first internet bubble of the mid-nineties was that we didn’t have the world’s attention on us back then. We could play around with emerging technologies on a weekly or monthly basis without anyone really noticing. Now, as Tom said, AI is helping us with everything from writing emails to brainstorming new ideas—and the whole world is watching. The question is, can we, as an industry, keep up with that pace?

Karim: Yeah, one area where I’ve found AI especially useful from a development perspective is rapid prototyping. For example, I mentioned building a settings page—if I just want to rapidly prototype a page to show a client, not necessarily focusing on the underlying code but on how it could look or function, I can do that in seconds with a simple prompt. That’s actually pretty powerful. Once I get client buy-in, then we can go back and build it the proper way—at scale and with enterprise-level standards. As you said, Brad, enterprise software has to check every box—scalability, security, performance—before it even gets to staging, let alone production. So, blindly trusting AI-generated code without verification is risky. I suspect some developers are already trying to hide how heavily they’re relying on AI-generated code, but at the enterprise level, that’s going to get caught quickly. Enterprise-level scrutiny means every line of code gets reviewed, and you can’t hide sloppy or insecure implementations.

Brad: Exactly. What you’re describing, Karim, is spot on. I’ve been surprised by how quickly some people jumped to the conclusion that AI would replace developers overnight. It’s true that it can assist with code generation, but it often lacks efficiency and nuance. The other thing I’ve noticed is how enterprises are reacting to AI. There’s a significant divide—some companies are pushing ahead and adopting AI aggressively, while others are more cautious, even banning its use due to concerns over data provenance, liability, and intellectual property. For example, we’ve had clients say, “No AI-generated code on our projects,” which puts us in the position of having to educate them on where and how we use AI—particularly in areas like QA, automation, and internal tooling—while assuring them that human review and adaptation remain critical. Internally, we’re dedicating a lot of time to testing new models and tools. It’s becoming a daily task—someone says, “Hey, check out this new model,” and we drop everything to evaluate it. Honestly, it’s adding a lot of internal costs. At some point, we’ll have to decide if this breakneck pace is sustainable or if we should switch to a more strategic evaluation cycle. Right now, the landscape changes so fast that you barely finish evaluating one tool before a newer version launches.

Tom: Yeah, that’s one of the fundamental challenges. The pace is so rapid that if you’re not continuously testing the latest tools, your assumptions quickly become outdated. It reminds me of conversations with people who tried ChatGPT 3.5 and found it clunky, then assumed that all AI models were at that level. But the technology has already leaped forward, and in three or six months, it will be in a completely different place again. It’s similar to what we saw in the early internet days—except now, the whole world is watching. That’s the difference—before, we could experiment without pressure. Now, the public and our clients expect results from these tools immediately.

Karim: One area that’s impressed me lately is the emergence of agent-based models. For example, I’ve been using OpenAI’s new “Deep Research” model—it’s the first agent I’ve found genuinely useful. You can prompt it with something like, “Produce a report on trends in the enterprise CMS space,” and it will spend 30 minutes searching the internet, then return with a comprehensive 5,000-word analysis. Over the past few days, I’ve generated 20 of these reports, and it’s already been hugely valuable for my strategic planning. In the past, I would’ve hired a researcher for that job—now, I just ask the agent.

Brad: Interesting—you’re using the Deep Research product from OpenAI?

Tom: Yeah, I’ve tried that one too. I’ve also tested Google’s version, which is confusingly also called “Deep Research,” but I found the Google version to be much more shallow. It lacked depth and didn’t produce nearly as comprehensive results.

Brad: The reason I ask is that I’m curious about how quickly people might start blindly accepting what these models produce without verifying the results. Internally, we’re telling our developers to “question everything.” We want them to use AI to accelerate their work but also to challenge its output—treat it as a tool to check their own thought processes, not a source of truth. If people start accepting AI-generated content without scrutiny, we’re going to see problems—especially with less-experienced developers. Sure, these tools can generate solutions quickly, but can they build correct solutions? That’s the difference.

Tom: Yeah, I think people will default to the path of least resistance because it’s easier. The question is whether AI improves fast enough to overcome its issues with hallucination and inaccuracies. The thing I’ve liked about the Deep Research tool, compared to just using GPT-4, is that it actually searches the web, pulls real-time information, and cites its sources. It’s not relying purely on training data. That makes it more transparent, and you can click through to verify the sources if something looks off. That’s a big step forward because it reduces the “black box” problem.

Karim: And imagine what happens when these agents can replace traditional user testing. Think about it—if a model can interact with a website, click around, and attempt to complete tasks you assign it, that’s powerful. But the question is—can it replicate human behavior or will it only behave like a bot? That’s a huge difference because if it doesn’t behave like a real user, then the data is useless. Still, if it gets good enough to simulate real human interactions, it could be an incredibly valuable tool for usability testing.

Brad: Exactly. Imagine a future where you combine AI with customer data platforms (CDPs). Instead of just analyzing user journeys after the fact, you could have AI-driven personalization in real time—customizing the user experience based on their behavior as they navigate your site. Hyper-personalization is one of those buzzwords, but the potential is real—especially for enterprises that want to tailor experiences to individual users.

Karim: Yeah, that’s the direction we’re heading—predictive UX, real-time personalization, and dynamic content tailored to the user. It’s both exciting and a little scary—especially when it knows more about me than I’d like! But for enterprises, the value is clear—if you can serve hyper-relevant experiences to each visitor, conversion rates will skyrocket. That said, there’s a privacy side to consider. I think enterprises will need strong policies on how they use that data to maintain user trust.

Tom: Right, and this all ties into what’s happening in the broader CMS and DXP (Digital Experience Platform) space. Personalization has been a major trend for years, but Karim, you were recently at a CMS industry conference—can you share what trends stood out to you?

Karim: Absolutely. In January, I attended the CMS Kickoff event in St. Petersburg, Florida. It’s an annual gathering hosted by Boye & Company that brings together leaders from both proprietary and open-source CMS platforms. It’s a unique event because there are no customers—just vendors, agencies, and ecosystem partners—so you get a lot of candid conversations. This year’s theme was AI and personalization—no surprises there—but there was also a strong focus on multi-CMS strategies. In 2020, 60% of enterprises reported using multiple CMS platforms—that number has now jumped to 85%. Enterprises are leaning into the “best tool for the job” mentality. For many, their legacy CMS isn’t going anywhere, but they’re using WordPress alongside it for rapid content delivery—landing pages, campaign microsites, blogs, and newsrooms. WordPress has become the perfect “secondary CMS” in these multi-platform environments because of its agility and ease of use. That’s where a lot of the growth is happening.

Tom: Yeah, we’re seeing the same thing. Enterprises that have massive, legacy platforms—too large or too entrenched to replace—are propping up WordPress next to them to handle fast, flexible content needs. It’s a great fit, especially with the block editor enabling rich, component-based content creation.

Karim: Exactly. And what’s interesting is that open source continues to gain traction in the enterprise space—not just for cost savings but for security and flexibility. The latest State of Open Source Report from Red Hat highlighted that security is now one of the top reasons enterprises choose open source. That’s a huge shift. Open source used to be seen as a security risk—now, it’s seen as a security advantage. But this also raises the stakes—because enterprises expect strong security practices from the open-source ecosystem.

Brad: That’s why I’ve been vocal about supply chain security in the WordPress ecosystem. Enterprise customers want transparency, signed packages, verified sources, and secure distribution strategies—things that are standard in other open-source ecosystems like Linux. We have an opportunity to lead here, but we need to act. And to be clear, I’m not advocating for any sort of fork or schism in the WordPress project—I’m advocating for a safer, more enterprise-ready distribution model. It’s about making sure that enterprises can confidently build on WordPress without worrying about supply chain vulnerabilities.

Tom: Completely agree. And I think it’s a positive sign that WordPress is finally part of these broader CMS and enterprise conversations. For a long time, WordPress dominated market share but wasn’t invited to the “enterprise table.” We’d go to these conferences, and people had outdated or misinformed assumptions about WordPress. But that’s changing. More enterprises are using WordPress for mission-critical projects, and the industry is taking notice.

Karim: I think part of the shift is that enterprises are seeing WordPress’s potential beyond just blogging. Features like Full Site Editing, the REST API, and the block editor have turned it into a true content platform. But it’s been a slower adoption curve—especially with Full Site Editing. Unlike the block editor, which enterprises adopted relatively quickly for content management, Full Site Editing requires a bigger workflow shift. Enterprises are cautious because they need guardrails—controls that prevent a junior marketer from accidentally breaking the global header.

Brad: Yeah, real-time collaboration and multilingual support are going to be the next major shifts. Real-time collaboration will be huge for media companies—imagine collaborative editing directly in WordPress instead of using Google Docs and copy-pasting. And multilingual capabilities in core, while controversial, are essential for global enterprises.

Tom: Absolutely. And it’s going to be fascinating to see how AI intersects with all these trends—especially real-time translation. Will AI translation models reduce the need for deeply integrated multilingual features, or will enterprises still want the control of managing localized content natively in WordPress?

Karim: Yeah, and as we’ve said—AI is advancing fast. Real-time translation used to be a gimmick—now it’s built into browsers and getting impressively accurate. Still, enterprises will want to control their messaging—especially for critical translations—so multilingual in core is still a huge step forward.

Brad: And enterprises will want AI-assisted translations they can customize—with their brand lexicon and industry-specific terminology. That’s something browser-level translation can’t offer.

Tom: Exactly. Well, we’ve covered a lot—AI, personalization, multi-CMS strategies, WordPress roadmap items—it’s been a great conversation.

Karim: Yeah, fantastic show—always good to catch up and hear your perspectives.

Brad: Absolutely—great to have the band back together!

Tom: Thanks, everyone, for listening—or watching if you’re joining us on YouTube. Be sure to check out past episodes if you haven’t already, and we’ll see you next time!