Advancements in Drupal AI and LLMs prompted us to explore ways to use AI to enhance and improve UX. Working with Drupal AI specialists, we spent a day considering how AI can impact user experiences of those publishing, searching and consuming University web content. 

Drupal is leading the way in applying AI to content management  

When it comes to current and emerging trends in content management, AI is top of the list. Drupal is an open-source content management system used by the University as well as many other Higher Education institutions worldwide. I became pertinently aware of the potential of AI for Drupal content management in my UX work contributing to the development of Drupal’s new non-developer-friendly product, Drupal CMS. AI features heavily in this new product, with AI Assistants and AI Agents being revealed and showcased at key points in the Drupal calendar.  

Learn more about Drupal AI developments from keynotes from Drupal’s founder Dries Buytaert:   

Driesnote, DrupalCon Singapore, December 2024  

Driesnote, DrupalCon Barcelona, September 2024  

Recognising the potential of this AI technology for the University, I worked with Jamie Abrahams, Drupal AI track lead and founder of Freely Give, the agency leading Drupal AI development to carry out research on the AI Assistants with University staff, to investigate the user experience (UX) they provided.   

Read about my work researching the UX of Drupal AI Assistants (December 2024) 

Making AI useful and usable – consolidated learnings from UX research of Drupal AI Assistants 

Leading on from this work, I was keen to work more closely with Freely Give to ideate around ways we could bring AI to aid content publishing and management at the University. 

Adopting AI in a controlled way to understand and explore its potential  

With any new technology, there’s a temptation to get onboard with it quickly, to ensure keeping up with the latest developments. Given the power, potential and costliness of AI, however, a more responsible approach is to implement a multi-phased AI adoption, including Exploration and Evaluation before Expansion (as described by Olivia Gambelin in her book ‘Responsible AI : Implement an Ethical Approach in your Organization’). The Exploration and Evaluation phases aim to grow understanding on the basis of careful experimentation. In alignment with the University’s Generative AI guiding principles, and with an appreciation of Drupal AI developments and functionality gained from the open-source Drupal community, I proposed working with Jamie and the team at Freely Give for a day to explore benefits of its potential application in the specific context of the University website.  

Read the Generative AI Guidance for Staff on the University website (published November 2024)  

For AI design to succeed, it requires a UX approach  

I had familiarised with some of Freely Give’s previous work (for example, automating rail claims) and from wider reading, had learned about the processes to design and apply AI to automate tasks traditionally completed by humans. A common theme when designing an AI solution was in line with other UX design approaches: To break down or map out the complete series of stages a person would go through to complete a task, work out which stages were the most difficult (either because they presented ‘pain points’ for the person, or because they were fiddly or prone to error) and then figure out which of these difficult stages could be done by an AI. Assessing whether the AI had made things better came down to whether the end result was acceptable and as good as if a human had completed the task, and crucially, if the AI had made things easier for the person.  

Jamie Abrahams: Fast Tracks to Compensation: Revolutionising Rail Claims with Drupal and AI – A Case Study   

Identifying areas of content management and design to apply AI to 

The first step in planning what to work on with AI was targeting specific web-related tasks that presented difficulties, either to people creating, editing and managing content (for example, University web publishers and site owners), or to people using web content (for example, end-users and audiences of the University websites). Several areas were well-known to the UX Service based on our previous research with web publishers and our knowledge of University website audiences. In order to ideate about AI opportunities, it was helpful to frame these areas as a shortlist of problems and actionable questions, which included:  

• How could AI help people apply the rules in the Editorial Style Guide?  

• How could AI-enhanced search improve the findability of content?  

• How could AI help people structure content into Drupal Paragraphs?  

These three problems provided us with something to work on during the AI day with Freely Give, however, we acknowledged that we may not be able to tackle all three issues in the time we had.  

Time for learning and time for experimentation  

In order to take an experimental approach, we needed to begin the day spending some time learning about Drupal AI, to understand the different solutions available and to appraise the relative usefulness of these in addressing our identified problems. Jamie provided an overview of key Drupal AI features including: AI Automators, AI Agents, AI Assistants and AI Search. To help us put the information we learned into practice with experiments, we were fortunate to be joined by AI developers Marcus Johansson, Andrew Belcher and Michal Gow as well as search specialist Charlie Hull.  

In the course of the day, we had time to work on two out of the three problems we had shortlisted, which were evaluating content against the Editorial Style Guide and experimenting with AI-assisted search. Details of what we did and what we learned are detailed in separate blog posts:  

Can AI help or hinder search? Trials with Drupal AI-boosted search and AI Assistants

An automated Editorial Style Guide? Experimenting with Drupal AI Automators