Drupal: Power, Flexibility, Freedom, and Now Smarter with AI
James Abrahams, Director at FreelyGive Ltd and the AI Track Lead for Drupal Starshot, has spent over a decade shaping the Drupal ecosystem, bridging the gap between complex technology and real-world usability. A consultant, software architect, and advocate for open-source collaboration, he has played a key role in making AI more accessible within Drupal. His latest work, the Drupal AI module, is a game-changer. It introduces an abstraction layer that allows seamless integration with multiple AI models, empowering site builders with little to no coding experience.
In this interview, James discusses the inspiration behind the AI module, the challenges of scaling FreelyGive, and the collaborative effort with key contributors like Marcus Johansson and Kevin Quillen. He also shares insights into his leadership role in Drupal Starshot, the evolution of AI in the CMS space, and the broader impact of AI on workflows, automation, and content management. His candid reflections on ADHD, productivity, and the highs and lows of innovation add a deeply personal dimension to his journey.
Speaking with Alka Elizabeth, sub-editor at The DropTimes, James offers a behind-the-scenes look at how AI is reshaping Drupal, the hurdles of driving cutting-edge technology in an open-source community, and what the future holds for AI-powered content management.
TDT [1]: You started your journey building a church website with phpBB at 16. How did that experience shape your career trajectory, particularly your interest in Drupal and CRM systems?
James Abrahams: I’ve always been deeply excited by online communities and the Internet’s ability to bring people together since I was about 12 years old, reading magazines about an online game called Ultima Online. Then Wikinomics radicalised me further, saying that mass collaboration would change everything. This book, combined with the forums I ran for gaming servers, led me to build online communities for churches.
We noticed that many of the random conversations became more structured. Some posts were really articles that could be tagged in a resource library of common questions about Christianity, while others organically turned into Events where people wanted to meet. So, we wanted to create a dynamic CMS where you could tag a forum post or thread, and it would suddenly become content for the website (such as appearing in a resource library or a calendar).
We created this with phpbb and thought about building it out into a product for churches. Whilst we were excited by online communities, we felt that CRM functionality, donation management, etc... would sell but give us an opportunity to build what we really cared about. We initially started with Joomla. We installed an event management system using CiviCRM Events in 2 days, but then a client asked us to add the field “Doors Opening time”, and that took us about 4 days to figure out! We quickly fell in love with Drupal’s CCK, and now fields and Views as a result of this!
Early on, we knew that we’d need a CRM system to enable all the other features of Church Management. I thought it would take us maybe 3 months to build, so we started an initiative with others at DrupalCon London 2011. Almost 15 years later, we’re still working on it!
TDT [2]: FreelyGive has grown from a small group of freelancers to a team handling diverse projects. What have been the key challenges and milestones in scaling the company, and how has your leadership evolved over time?
James Abrahams: The biggest challenge hasn’t just been due to pure scaling but the unique nature of AI and being in an ever-changing environment. AI is constantly changing and so the sales targets have been constantly moving. It’s also hard to know if something will work until you build it, putting a lot of pressure on working prototypes over nice ideas.
This has led to an incredibly fast and constantly changing way of working, which is great for most developers’ way of working. However, It also means everyone involved must learn a ton of stuff before getting started.
We’re working on reducing this difficulty internally but also reducing this difficulty across the AI community with better documentation, more planning, clearer roadmaps and issues. But it's difficult to provide these things when the whole world is scrambling to do something useful.

TDT [3]: You’ve mentioned that 2024 was transformative for your journey into AI. What excites you most about the potential for AI?
James Abrahams: As a non-developer, I’m really excited by the productive power AI can bring to people who, like me, weren’t able to do a lot of this stuff ourselves. Being free to be creative and get ideas working without having to go through another person feels amazing. Developers will always be necessary to do ever more complicated tasks. But AI has already enabled me to work faster and remove so many irritating parts of my job (like converting a random list of notes into a spreadsheet with hours estimated and calculated!)
I’m excited to see human potential unlocked as AI allows us to automate things in an easy-to-understand and transparent way. So many areas of life drag because our processes are slow (the backlog of courts in the UK is awful at the moment), and so I’m looking forward to a time when most of the things we want can be done quickly!
TDT [4]: What inspired the abstraction layer approach in the AI module, and how has it improved AI integration in Drupal?
James Abrahams: I think it was inevitable; the question was more about when, not if. We LOVE flexibility in Drupal. One problem, but also the strength of Drupal, is you can always do the same things in so many ways. Drupal is about having freedom. We have database abstraction layers; we’re REST first because we don’t care where you have your data; we’ve supported so many different front-end frameworks. We always knew at some point, we’d need an AI abstraction layer. In 2023, it was all about OpenAI, but other models started catching up, and even for our internal projects, we wanted to know how good they were—so we could compare them.
I actually pushed back on the abstraction layer for some time. My view is that I wanted to try and build something useful that people would pay for first before worrying too much about building a structure that is “Correct”, but when Kevin Quillen suggested we build a new AI module with that namespace, the time felt right. We’re now seeing so many real-world applications of AI that produce real value and we’ve moved beyond just making demos.

TDT [5]: What was the initial idea or challenge that led to the development of the Drupal AI module? How did you envision it addressing gaps in the Drupal ecosystem?
James Abrahams: We already had built a claims management system in Drupal. Early on, we saw the power of using AI to help automate those claims. We saw how powerful it would be for sales if we built reports evaluating the accuracy and effectiveness of AI by comparing AI-generated answers to human-evaluated ones. Putting the power to “prompt engineer” (try things out and change the instructions to AI) in my hands, as a non-engineer, felt amazing.
Early on, I always felt that tools like LangChain or the llama index, which are so focused on developers, would prevent many people like myself from benefiting from AI. So, I felt AI needed Drupal.
In order to sell it to the Drupal community, we needed to find out why Drupal needed AI. So, we tried to tackle what most people find the hardest problem with Drupal: its usability. Marcus Johansson built this amazing demo of using AI agents to create a webform by drawing it. We showed it to community members, and gradually, it morphed into what we have today in Drupal CMS.
Drupal is powerful, flexible, and gives users freedom. With AI, it can also be a joy to use.
TDT [6]: The AI module is tailored for site builders with little to no coding experience. How did you balance technical sophistication with user-friendly design during development?
James Abrahams: It’s primarily because I can’t code that I’ve been motivated. As an agency owner, if someone builds something in AI for drush, command line or a full black-box application using one of the Python frameworks, I can’t know what is happening or improve it myself. So, selfishly, I pushed for many of these things because I wanted them for myself. Marcus is a developer but has also tried hard to make tools non-developers could use with his Mini Kanban and AI Interpolator (Now AI Automators), so I really wanted to work with him.
One thing that makes it easy is that AI itself does a lot of the heavy lifting, and the whole purpose is that you can interact with it in a natural language. I studied physics with philosophy, but the philosophy of language that I studied has been instrumental to a lot of my work in systems architecture and turning ideas from clients into tasks, and it really helps with AI. I think people with psychology degrees, management degrees, and English literature degrees are going to be far more important than they were when creating AI-driven systems.
So, to some degree, AI isn’t as technical as previous technologies, which helps with making it easier to use.

TDT [7]: The module has been a collaborative effort involving key figures like Marcus Johansson and Kevin Quillen. Could you share insights into the teamwork dynamics that drove its success?
James Abrahams: I knew from the beginning that AI should be bigger than one company or organisation. It has the ability to touch almost every aspect of our lives, so even as early as 2023, I wanted to build a community around AI rather than just build things myself. This kind of goes back to the days of Ultima Online and getting into online communities. So, I spent a lot of time reaching out to people like Kevin, Marcus, or anyone actively doing things with AI.
All the maintainers of the AI module felt similarly and tried to do things publicly rather than behind closed doors for specific clients. This has been great because everyone on the AI maintainer team has different focuses and areas of expertise. This has driven success, as people have been able to work independently from their various perspectives.
I pushed back on AI abstraction layers, but Michal Gow went and built the LLM provider anyway. So when we needed it, Marcus could adapt it to the core of the AI module. I pushed back on Scott Euser from Soapbox’s “Boost with AI”. Doing vector search to make views better with no LLMs seemed boring to me, but he saw an immediate need for it with his clients, and it's been truly amazing to get people into AI without many of the risks.
So, I think it's been powerful to work on things independently but within a common framework that can share the benefits.

TDT [8]: As the lead of the AI track for Drupal Starshot, how did you approach identifying and prioritizing the AI features that would have the most significant impact on Drupal users, particularly non-technical site builders?
James Abrahams: I’ve worked very closely and collaboratively with the Drupal CMS leadership. I’m so immersed in AI, and I care about it so passionately that it's difficult for me to know what is “cool” for other people. So I’ve spent most of my time exploring the art of the possible, presenting things to the rest of the leadership, and then seeing what sticks.
One thing I did bring to the team was the purpose of Drupal CMS. Everyone was talking about why Drupal CMS needed to exist because Drupal had a reputation for being difficult to use. We had graphs showing how people loved using Drupal after a period of time and hated most of our competitors over that long time period, but initially, most people struggled with it.
For Marcus and me, those immersed in AI, it became clear as day that AI would take this even further.
TDT [9]: With the first version of Drupal Starshot now released, what are your long-term goals for the AI track, and how do you envision AI continuing to transform Drupal’s capabilities and user experience in the years ahead?
James Abrahams: I think there are two streams we think will be important in the medium term, continuing the trend of focusing on what people find painful and making it easier. For developers and site builders, that’s migrations. Migrating data, code, modules, concepts, config, refactoring. All of this is horrible. The move from Drupal 7 to Drupal 8 was so painful it almost discouraged Drupal from making such sweeping changes. When you spend a year migrating a client, you’re spending that year working and spending a client’s money without them seeing any value until it is done.
If we can make it quicker, if we can easily migrate even the designs into Drupal, I think this will propel Drupal CMS even further and make it much less painful for us all to innovate.
The second stream is allowing non-technical end users to work the way they want and still benefit from Drupal. So, tools that allow people to work in design tools like Figma and then use AI to create a theme, create content in a Google doc, and then have it migrated into the structure of Paragraphs
and experience builder
(XB) will be powerful in improving the content editing experience.
I’m still all about the native Drupal CRM and automation, so I’ll be plugging away at that behind the scenes. I don’t talk about it in Drupal, as most of the problems I try to solve and the clients I want to win are different from what Drupal is used to.

TDT [10]: What were the biggest technical or community-related challenges you faced while developing the AI module, and how did you address them?
James Abrahams: That question is in the past tense, and I think the biggest challenges are ahead of us! We haven’t addressed them; we’re just solving them over and over. The biggest two challenges right now are how to make the pace of innovation sustainable. It’s been exhausting over the last two months and although there was a lot of interest and LinkedIn posts, that didn’t translate into income. This is turning, and I believe 2025 is the year many clients will be doing real paid-for AI projects to solve immediate problems, but we need to make this financially sustainable and sustainable for our mental health!
The next big challenge we’re working on is trying to bring more people into the community. A lot of the stuff we did was just me and Marcus plugging away for Drupal CMS because we had to move quickly to the changing needs of the Drupal CMS leadership and then make these AI Agents work. The structure of how AI Agents work is far from decided in the AI community, and so we had to figure this stuff out as we went along, which meant it was hard to bring someone in who wasn’t as obsessed with AI as we were.
The change in AI structures has slowed a little recently, so we’re able to focus on taking what we’ve done and building useful things with it. We’re also really trying to figure out how to organise a whole community of individuals and agencies around AI.
TDT [11]: With the release of the AI module 1.0 and plans for DrupalCon Atlanta, what are your aspirations for the module and AI’s role in Drupal over the next few years?
James Abrahams: I think this grows bigger than me, and I’ll be able to focus back on AI in Drupal CRM and automating specific things. I just love workflows and business processes and how people communicate internally to make something happen. I’ve always been less of a marketing or content person and someone who wants to help people create tangible things more effectively. At the moment, I’m riding out this wave and letting it see where it goes.

TDT [12]: You’ve been open about how your ADHD diagnosis and Vyvanse treatment improved your productivity. How did this newfound clarity shape your contributions to the Drupal AI module and your work with FreelyGive?
James Abrahams: It changed everything. None of this would have happened without the diagnosis and treatment with medication. I tried so many things and struggled with them; maybe I’m lazy and have poor self-esteem. People would compliment me, say I was clever, and congratulate me for starting a company when I was young, but I would have many weeks doing next to nothing, finding minor client drama paralysing if I had to send an e-mail.
I remember knowing that before I went to a Drupal event, I needed a video showing off these agents. It had to feel like a Tiktok; it had to be 1 minute. I had to show something super cool, super quickly, so I could show it to someone in a pub. Marcus made a video that was 10 minutes long. Old I would have criticised it, maybe told the dev to go and try and make it shorter.
But instead, I just started writing, I watched it over and over and took out all the best bits from that video. I thought about the bits that impressed me the most. I wrote a script for it, used AI to make an audio version of my script, and then timed it. I got it down to 1 minute and then could give it back to Marcus to edit.
Now, this wasn’t perfect. Now, I can edit videos myself, and it was one of the first videos I did like that, so it wasn’t the greatest thing ever created. However, I needed that video for that evening, and from that evening, a lot of interest started. This happened maybe two weeks after I started getting treatment. I don’t think any of this would have happened otherwise.
Now that I’m more productive, the next step is to be able to chill out and not put all my self-worth into success with the AI module! My personal life is fine; my wife loves her job, and I love looking after my kids, so we’ll be fine. It's SOOO much fun, but I have to be careful not to overdo it because of my ego.
TDT [13]: On an ending note, looking back, what has been the most rewarding moment in your journey with Drupal?
James Abrahams: Since getting into AI, I’ve met some amazing people, especially the Indian Drupal community during DrupalCon Singapore. It's been awesome to watch their passion and excitement about what we’re doing with AI.
For fun, I watch YouTube videos and streamers. So, people creating reaction videos or tutorial videos trying out the AI agents has been WAY more rewarding than I thought it would be, or perhaps it should be
Ivan Zugec made this video.
He got the Drupal AI Assistant to do something but forgot to add things to his prompt. So, in what looked like a live video, he tried some new prompts. These were things I had never tried myself, so I was on the edge of my seat watching this video. Was it going to work?!?!?!?! They did, and he kind of let out this truly happy and surprised laugh. I’ve had that once before with the CRM work I’ve done, and I love those moments.