London, February 24 AI could soon better understand people's accents and dialects, thanks to a landmark study being led by the University of Sheffield.
The study, led by Dr. Chris Montgomery from the University's School of English in collaboration with ICS.AI, the UK's fastest-growing, profitable AI business, is set to tackle one of the biggest challenges in public sector AI – how well digital services understand people who speak with different regional accents and dialects.
The collaboration is the first UK academic-industry partnership to apply sociolinguistic research to the evaluation of public sector conversational AI, focusing on how systems perform in real service interactions between citizens and public bodies - for example, when people contact their local council and conversational AI is used to answer questions and help them access the right information or services.
The study is based on Dr. Montgomery's systematic review of more than five decades of peer-reviewed research on accent and dialect variation across Great Britain.
This follows a recent survey, which revealed that over half (52 per cent) of UK residents are concerned that AI may struggle to understand accents or dialects. This increases to 57 per cent in Wales, 67 per cent in Northern Ireland, and 71 per cent in Scotland.
A Literature Review, which started ICS.AI’s collaboration with the University of Sheffield in December 2025, found that misunderstanding and bias are most likely to arise not because of how people speak, but because of how speech is interpreted.
While most existing studies focus on speech patterns, far less research examines how listeners recognise and judge accents and dialects in practice.
The findings provide an important evidence base for how conversational AI should be evaluated in real public-service settings, helping to ensure that systems are tested in ways that better reflect real-world variation in speech and communication.
Dr. Chris Montgomery, Senior Lecturer in Dialectology at the University of Sheffield, said: "Many people across the UK may have found themselves in a situation where they've tried to phone their local council - perhaps because their bins haven't been collected or they've had an issue with their council tax - but the AI that first handles their call struggles to understand them."
"As AI is frequently used to direct calls across public services, we need to ensure the technology can understand the range of accents and dialects it may be faced with," he added.
"This project brings sociolinguistic theory and evidence on listener behaviour into applied evaluation, enabling performance claims to be framed in ways that are both scientifically defensible and socially meaningful," he said.
Dr. Crispin Bloomfield, Chief Education Solutions Officer at ICS.AI, added: "Public sector AI has to work for everyone, not just for people whose voices or speech patterns are easiest for systems to process. This collaboration empowers ICS.AI to apply established sociolinguistic evidence directly to how conversational AI is evaluated in live public service environments, helping us build inclusivity in a transparent and scientifically grounded way."
ICS.AI and the University of Sheffield will work together to translate their findings into practical evaluation frameworks and new capabilities within the ICS.AI platform. In parallel, the University of Sheffield will lead further research into misrepresented dialects to strengthen the evidence base that will underpin future joint work with ICS.AI.
The collaboration reflects the University of Sheffield’s commitment to independent thinking and a shared ambition, demonstrating how creative minds at Sheffield are shaping solutions to national and global challenges.
