There's no shortage of conversation about artificial intelligence. But we're more interested in getting our hands dirty and finding out what actually works.

Recently, we brought our Drupal team together to share the AI tools, techniques and experiments they've been exploring.

There were live demonstrations, some clever ideas and a few moments of "Wait, you can do that?"

Here's a little look behind the scenes.

A new addition to the developer's toolkit

Our developers are experimenting with AI-assisted coding tools that work directly with Drupal project codebases.

Rather than simply asking AI a question and copying the answer, these tools can understand existing project structures, suggest changes and help develop new functionality within the context of a website.

During one of our demonstrations, a developer used AI to help build a custom Drupal feature for exporting taxonomy data into a CSV file.

The AI explored the existing codebase, generated the necessary functionality and presented its changes for review.

Pretty clever.

Of course, the important part is what happens next. Our developers still need to inspect the code, test the functionality and make sure everything works as intended.

AI is a useful assistant, but it doesn't replace the experience needed to build reliable, secure websites.

Making Drupal maintenance smarter

Keeping Drupal websites secure and up to date is an important part of what we do.

We're exploring how AI can support our existing maintenance processes, alongside automated updates, testing and deployment workflows.

One of our developers has been experimenting with Bitbucket pipelines that regularly check for Drupal updates, apply changes and use AI to help review the results.

The idea is to make maintenance more proactive, with smaller, more manageable updates and clearer visibility of anything that needs attention.

For organisations relying on Drupal to deliver essential information and services, maintaining a secure, stable platform is critical.

And while automation can support that process, human oversight and proper testing remain essential.

Helping developers solve the tricky stuff

Not every development challenge comes with a straightforward answer.

Sometimes it's understanding a complex codebase, investigating an unexpected issue or working through a particularly detailed technical requirement.

We're using AI to help our developers explore code, investigate problems, review documentation and consider different approaches.

It's a bit like having another set of eyes on a problem. One that can process a lot of information, but still needs an experienced developer to decide what makes sense.

The benefit is having more tools available to support good technical decisions.

And ultimately, that's what matters to our clients.

Sharing ideas, not just experimenting in isolation

One of the most interesting things about our recent AI session wasn't any single tool or demonstration.

It was seeing how differently people across the team are using AI.

From developers experimenting with coding assistants to our UX team using AI to analyse research findings, everyone is finding different applications.

Bringing those ideas together gives us an opportunity to learn from one another, challenge assumptions and identify approaches worth exploring further.

We're now building more regular AI knowledge-sharing into our internal processes, with a focus on practical demonstrations, documentation and testing.

Because an interesting experiment becomes much more valuable when the whole team can learn from it.

Keeping people in the picture

As exciting as these tools are, we're not handing over the keys to AI.

It can make mistakes, misunderstand requirements and produce code that looks convincing but isn't quite right.

That's why our developers remain responsible for reviewing, testing and validating the work.

We're also mindful of how AI tools interact with client information, project codebases and security requirements.

For us, adopting AI responsibly is just as important as exploring what it can do.

What's next?

We're continuing to investigate how AI can support Drupal development, maintenance, testing and the wider digital delivery process.

Some of these experiments may become part of our standard workflows. Others could inspire new capabilities for the websites we build and support.

And some will probably end up in the "interesting, but not quite there yet" pile.

That's the nature of innovation.

We're not exploring AI because we want to replace the people behind our work. We're exploring it because we want to give talented people better tools to do great things.

And we're only just getting started.