Investigating the Hype: First Impressions of the Drupal AI Framework

AI is a broad field that spans many disciplines, but in recent years, it has become a buzzword. Does the general public truly understand what it means? Terms like "vectors" and "RAG" might sound impressive, but if end users are not able to follow along, what’s the point of all the hype?

Square360 decided to try out the Drupal AI framework and its submodules, unpacking some interesting developments that could impact CMS users. As we explored the modules, one test stood out. We set up a nickname field on user profiles and configured an AI automation to generate playful names based on each team member's bio. The automation named me, the business analyst, "Curious Maven," and gave our investigative developer the title "Bridge Master Doug." The naming felt surprisingly accurate, so in the spirit of AI, we're using those names in this interview.

Read on as we discuss our initial thoughts on the Drupal AI module.

Curious Maven: So, broadly speaking, how has Square360 been approaching AI?

Bridge Master Doug: I think with any new technology—especially something like AI, which is surrounded by a lot of hype—we like to approach it with the mindset of an "open-minded cynic". That means being curious and receptive, but also critical enough to cut through the noise and ask the ultimate question: Will this actually bring value—to our clients, and to us?

AI is easy to get excited about. The possibilities it opens up are incredible. But at the same time, the cynic in us can just as easily point out its flaws—and there are plenty. Klarna recently reported significant losses after replacing a large part of their workforce with AI. A recent experiment by researchers at Carnegie Mellon ran a simulation using AI to staff an entire company and the results were very sobering. It's far from perfect. It often produces inconsistent or incorrect results. But that doesn’t mean it lacks value.

At Square360, we’ve been gradually integrating AI into various areas—not just in development, but in project management and beyond. Ultimately, it's a tool. Like any other, it has strengths and weaknesses. The more we understand how it works and the more we use and experiment with it, the better positioned we are to get real, tangible benefits.

Curious Maven: The Drupal AI module has been out for about a year now. What were your first impressions when you started exploring it?

Bridge Master Doug: I was really surprised at how advanced the module’s ecosystem had become in what seemed like such a short space of time. There are lots of features and integrations. It’s very well thought out and extensible. I’d say that’s point number one. Point number two is that the community around this ecosystem is very active, open, and helpful. Even though development is happening very quickly, there are lots of videos being put out on YouTube regularly, detailing the changes and plans for the future.

Curious Maven: Were there any surprises, like features you did not expect or assumptions that were challenged?

Bridge Master Doug: I think the biggest surprise was seeing a demonstration where someone drew a form on a piece of paper, uploaded a picture of it to Drupal, and asked Drupal to generate a web form based on the drawing. At the time, that blew my mind. I wasn't expecting that level of automation so quickly.

Curious Maven: You mentioned to the team that you could set the AI’s tone using specific characters like a pirate or dialects such as Scottish or Cockney. How did that work in practice, and what kind of output did you get?

Bridge Master Doug: This is part of the way I like to experiment, alongside maybe a real-world example for our clients. I also like to do a more playful experiment to push boundaries. Many clients want to make sure the content they publish to the world follows a tone consistent with their brand, which might be formal or familiar. Now, I don’t think any of our clients have pirates as part of their brand identity, but again, it’s just a fun way to work, push the technology to its limits, and see what happens.

The practical applications emerging from these early experiments include processors that assist authors in preserving distinct tones for different contexts and filters designed to block inappropriate content. For example, if we’re generating content for children, we’d want a filter or processor to ensure that even if we use a pirate tone, the language is appropriate for the age group.

Curious Maven: From a technical perspective, how does the module integrate with the rest of the Drupal ecosystem?

Bridge Master Doug: It is very well organized, making extensive use of Drupal’s plugin system. AI providers, Agents are all plugins so new versions can be created easily. There is an abstraction layer between applications and the providers so in theory, you are not tied to any single AI provider and can swap between Claude or ChatGPT for example and you’re application of chat bot would not necessarily have to change. Field automation means we can use AI directly in our existing entities and these can be chained together to build really complex and powerful processes with relative ease.

Curious Maven: How stable or production-ready does it feel at this stage?

Bridge Master Doug: The module itself I'd say is absolutely ready for production. Some features we found are virtually plug and play. You enable them and they just work. That said, I would still test thoroughly to make sure any generated output fits the client’s needs. Perhaps in the area of site configuration and automation we could see a significant shift in the future. Integration of the Event-Condition-Action module (ECA) integration is in the pipeline. That’s something to keep an eye on, as it could open up lots of different possibilities for content enrichment and automation.

Curious Maven: Did anything about the module feel more experimental than practical?

Bridge Master Doug: I can’t speak for all the submodules, but I see no reason not to start rolling these features out to a production environment. That said, each feature should be evaluated individually, especially any custom ones.

Curious Maven: What do you think the biggest learning curve will be for teams trying it out for the first time?

Bridge Master Doug: I think it will be easy for teams to jump in and start experimenting. That said, more complex tasks and features definitely require more careful testing. Teams will need to experiment thoughtfully to develop effective prompts for their specific use cases. AI responses can be inconsistent or occasionally inaccurate, so iteration and validation are key. However, for simple tasks the barrier is low.

To give one example, we’ve been experimenting internally with using AI to correct semantic HTML. Authors don’t always create HTML that meets developer standards, so we wanted to see if AI could fix that and maybe apply microdata. Our initial results were promising but inconsistent. There’s a fair amount of experimentation needed to figure out how to give AI effective instructions. Sometimes that means being very specific in the prompts. Other times, we have to lower our expectations and remove some features altogether.

Curious Maven: Do you see it supporting real editorial workflows at this point, or is it more useful for prototyping and exploration?

Bridge Master Doug: Absolutely. It depends on each client and their needs, but we are already rolling out select features. It's worth pointing out that the CKEditor plugin workflow for content authors is not 100 percent automated. The author selects some or all of the text and requests modifications by AI. The AI’s suggested result is then presented to the author, who can review and edit it before applying changes to the document. I think this workflow is really important to avoid mistakes or low quality output (AI slop). With that in mind, I think many features are ready to go right now.

Curious Maven: If someone asked whether Drupal AI is worth trying, what would you say?

Bridge Master Doug: Absolutely. Even if you’re a cynic, it’s important to be an open-minded cynic. Don’t assume AI is here to replace you, and don’t let its errors or faults discourage you. AI is an incredibly useful tool, and whether you’re a developer or a content creator, it can empower you. You’ll get out of it what you put into it. The key is to experiment and find how it works best for you. The only way to do that is to jump in and get your hands dirty.

Curious Maven: Looking ahead, how do you think this tool could change the way teams approach content and site management?

Bridge Master Doug: AI empowers everyone to some degree in different ways. I feel it’s far from perfect. The internet is now full of AI-generated content. Most of it is arguably low-quality and most of that we can probably spot, even though it keeps getting better. I believe the best results happen when AI works with us. For example, it can help improve or enrich our content, automate certain tasks. As a developer I experienced a significant increase in productivity using AI every day. Knowing it isn't perfect and makes mistakes is a key part of that - I carefully craft my questions and review everything it produces.

I hope AI will increase the quality and productivity for everyone in these areas. In terms of site management and configuration, it could help site builders without coding experience handle more complex tasks. From my experience, I’ve seen more people showing an interest in Drupal because of its powerful AI system. That will help Drupal itself grow. The more people join, the stronger the community becomes, which keeps it advancing.

Our Reflections

The Drupal AI module has already sparked curiosity and discussion within our team. Its flexibility opens the door to new ways of thinking about content workflows, user experience, and team collaboration. Our work with the module so far has shown enough to suggest this tool has real potential. Having introduced some content-related features here, we’ll take a deeper look in our next story at how the module integrates with editorial workflows and supports content authors.