What was the challenge?

Coleg y Cymoedd faced growing pressure on admissions and learner services teams due to high volumes of enquiries, many of which were out of standard office hours. Traditional channels such as phone lines and emails were either time-limited or too resource intensive, while website frequently asked questions (FAQs) were often unable to provide responses of sufficient quality and accuracy, creating further pressure on telephone enquiry handling. This was particularly prevalent during the lead up to the new academic year, a period where the need to improve recruitment conversion and learner retention was key. 

Who would benefit and how?

The intention was for prospective learners and existing students to experience improved access to accurate, multi-lingual information at any time of day, reducing friction during recruitment and enrolment periods. Admissions and Learner Services staff would benefit from reduced call and email volumes, enabling them to focus on higher value interactions and complex cases, while the college would gain improved insights into learner behaviours and demand patterns. 

What was the solution?

The college developed ‘Arny’, a bilingual AI smart assistant deployed on the college website. The project was funded through the Welsh Government Digital 2030 Capital Fund, which supported development and delivery despite limited internal capacity. A knowledge base of over 400 FAQs was compiled by staff from multiple departments, with answers provided in English and Welsh to meet language requirements. At launch, ‘Arny’ was a static AI chatbot able to provide answers or suggestions based on the knowledge-base alone, with separate versions of the tool available in English and Welsh, but it has since migrated to a generative-AI platform which can analyse wider data sets in order to provide more detailed, accurate support in a broader range of languages.   

Who was involved?

The project was initially led by the Digital Strategy and Innovation team, working closely with Admissions, Learner Services, IT, Technology Enhanced Learning and the Welsh Language department. Approximately 40 staff contributed to the knowledge base behind the first iteration of ‘Arny’, ensuring comprehensive coverage of an extensive range of common learner queries.  

Staff across departments were involved in content creation, testing and ongoing review, building organisational capability in digital service design and AI enabled support, while the Welsh Language department validated the accuracy and appropriateness of bilingual responses and ensured compliance with language standards. Regular reviews with technology partners and internal stakeholders ensured that the project leveraged external expertise and allowed for continuous improvement and shared learning across teams.  

What was achieved?

While direct causation has been difficult to prove, strong correlation has been observed between deployment of 'Arny’ and improved organisational outcomes, supported by quantitative metrics and qualitative staff feedback. Engagement and usage are monitored through Power BI dashboards which show clear seasonal demand patterns, such as usage spikes during peak recruitment periods and a shift in peak enrolment timing; with more enquiries being resolved more quickly, prospective students are able to make decisions sooner which, in turn, gives them more time to plan and prepare themselves for this critical next life-stage.  

'Arny’ handles over 10,000 queries per year, contributing to improved application-to-enrolment conversion, increased learner numbers and reduced attrition of new starters during their first academic term. The Admissions team have experienced a substantial reduction in phone calls and emails, with ‘Arny’s’ 24/7 availability improving responsiveness during peak recruitment periods and allowing the team to focus on more complex requests. The project has also been seen to support broader cultural change towards experimentation and digital confidence. 

How was the Welsh language considered?

Welsh language provision was integral to the project. All initial knowledge-base content was authored bilingually and the Welsh Language department validated response accuracy and appropriateness. Delivering a fully bilingual solution presented technical challenges, leading initially to separate English and Welsh instances of 'Arny’, increasing cost and complexity. This has been addressed by 'Arny’s’ recent move to a generative-AI platform which includes multilingual capabilities and real-time translation, meeting statutory requirements for Welsh Language support and benefiting a wide range of ESOL learners. 

What were the key governance steps taken?

The most critical governance factor was senior leadership support, including the creation of a Director of Digital Strategy and Innovation role and strong backing from the Principal, which facilitated a culture of risk-controlled experimentation. This was further supported by choosing to work with established suppliers who understood data protection and public‑sector risk, with priority placed on EU data residency and security controls. 

Before the first iteration of ‘Arny’ went live, user acceptance testing was undertaken over a two-week period, followed by monthly reviews to address gaps and improve performance, with data accuracy and language quality as central considerations. Ongoing project governance included cross departmental oversight and monthly supplier reviews, with lessons from the initial iteration informing subsequent procurement and platform transition decisions.  

The move to a generative‑AI platform was heavily supported by the supplier’s licensing agreement with JISC, which significantly simplified governance while reducing costs. This allowed Coleg y Cymoedd to run a three‑month pilot at a relatively low cost, reducing the initial financial risk and, in the longer term, making the solution more viable and sustainable within a constrained FE funding environment. 

What lessons have been learned?

Key lessons include the importance of senior leaders building their understanding of AI and its capabilities, initially in order to envision how it might empower people and accelerate strategic plans, then to be able to drive projects forwards. Alongside this, organisations need to create the roles, skills and capabilities that will enable them to horizon scan, lead innovation through experimentation and develop AI literacy. Finally, organisations should try to identify 2-3 potentially high impact starting points or use cases and start trialling ASAP, developing knowledge and experience while building the appropriate infrastructure and financial resources. 

Supporting information

Discover more about Coleg y Cymoedd’s approach to AI tools here