A useful customer-service AI system should help someone get the right answer, resolve a problem or make a better decision. Speed matters, but so do accuracy, privacy and knowing when a person needs to take over.
The Best Application of AI in Consumer and Customer Service category at AI Awards 2026 brings together Headon Boxing Academy, Laya Healthcare, Ministry Brands Ireland and SSE Airtricity. Their shortlisted work illustrates different ways of supporting customers and the teams serving them. View the official finalists.
Category sponsor: This award is supported by Rinn AI, the newly established research centre for data science and artificial intelligence.
Meet the four finalists
The following summaries draw on the published shortlist descriptions. They describe the applications; they are not independent verification of performance.
Headon Boxing Academy
Headon Boxing Academy’s Pulse identifies membership behaviour that may indicate cancellation risk. Its description combines prediction with staff outreach to support retention.
Laya Healthcare
Laya Healthcare’s Knowledge Hub supports advisers with policy-specific information, member context, references and confidence indicators to inform customer conversations.
Ministry Brands Ireland
Ministry Brands Ireland’s Amplify Intelligence lets churches and non-profits query information across systems in natural language, with access and approval controls.
SSE Airtricity
SSE Airtricity’s Enhanced Smart Insights Energy Advisor combines generative AI and smart-meter information to provide personalised guidance, with evaluation of accuracy, tone and relevance.
Read the full published category descriptions.
Four different ways to assess customer value
Our editorial view is that these applications need different tests. A retention tool, an adviser assistant, an information service and a guidance tool should each be judged against the customer problem they are intended to address.
For retention support, test whether outreach is useful. A prediction is only a signal. A sensible pilot would examine false alarms, whether staff can act on the information, and whether customers find the resulting contact helpful. Compare outcomes with a suitable baseline before attributing a change in retention to AI.
For adviser support, test the quality of the answer. Measure whether staff receive accurate, current information and can check the supporting source. Include ambiguous questions, conflicting records and cases where the appropriate response is to ask for help. Faster handling is valuable only when it does not create avoidable follow-up work.
For access to organisational information, test permissions as well as usefulness. A natural-language interface can be convenient, but evaluation should include whether each user sees only the information they are entitled to access. Test realistic questions and record when the system cannot provide a reliable answer.
For personalised guidance, test whether people can understand and use it. Review factual accuracy, clarity and the assumptions behind recommendations. A useful evaluation would distinguish between someone reading advice, acting on it and achieving a measurable result.
These are proposed evaluation criteria, not claims about results achieved by the finalists.
Start with a customer problem your team can measure
Before choosing a tool, identify one point of friction in the customer journey. It might be repeated explanations, difficult information searches or a handover that loses context.
Record what happens today. Then agree how the team will assess the pilot:
- Resolution: did the customer’s problem get resolved, or simply move to another channel?
- Accuracy: was the answer correct and supported by the right information?
- Effort: how much work remained for the customer and the staff member?
- Escalation: could someone reach a person when the system could not help?
Review difficult cases alongside successful ones. A short average response time can hide repeated contacts, corrections and customers who abandon the process. Give staff a practical way to report these problems and use that feedback to improve the service.
Recognition and evidence
Shortlisting is a reason to explore an application further. It does not replace evaluation in the setting where it will be used. Buyers should request evidence relevant to their customers, data and operating conditions.
Editorial disclosure: AI Ireland organises the AI Awards. This article uses published finalist summaries alongside AI Ireland’s editorial analysis. Shortlisting is not independent verification of performance claims.
Continue the series with our Agentic AI finalists article.
See practical AI in action
Join us at AI Awards 2026 to celebrate the organisations and people putting AI to work.
Want to bring these lessons to your organisation?
Discover more from AI Ireland
Subscribe to get the latest posts sent to your email.
