Title

It was 10 years ago this summer that I spotted a spate of news stories about Enfield LBC employing a virtual agent for its frontline customer services.Amelia was more than a chatbot. Its makers had based her on an American swimwear model called Lauren Hayes – blonde hair, blue eyes and a tailored business suit.They claimed she was fluent in natural conversation, emotionally engaged and had understanding of context. She could hold thousands of conversations at a time in multiple languages.Most major national newspapers covered the story: Council takes on its first robot worker (FT), Robot called Amelia to do the job of human council workers for the first time (The Telegraph) and The first public sector body in the world to recruit a virtual employee for frontline services (The Times).The original press release made no mention of the ‘R' word, yet almost all the press coverage framed Amelia as a robot – which is no surprise given the enduring fear-fascination about the idea of robots taking over.There was much about the cost savings, the convenience for customers, the point that Amelia would not deviate from the script, discriminate, bend the rules, strike or get ill. These are all familiar types of claims in the world of digital transformation – cost-convenience-control.But it was a fourth set of claims that stood out for me – the focus on Amelia's emotional intelligence to truly understand people's situation and how she could alter her facial expressions to show empathy. It was this idea of the perfect public servant that led me to start working on plans for a book project, and one I later published in 2020.By the summer of 2017, a year after the first announcement, there was still no sign of Amelia on the council website. It was revealed that the animated avatar had been dropped to focus on a live chat version and rather than general customer service, to specialise on handling planning and building control enquiries.By spring 2018, a new look Amelia was unveiled with a cropped hairstyle – said to be almost ready to go live to answer questions about 32 different types of planning matters.The executive director for resources capped off his presentation with: ‘Amelia doesn't smile any more, but we can absolutely guarantee she'll give you the right answer' . Yet it seems that soon after, Amelia was quietly shelved.The empathetic, cheaper, more reliable, intelligent, always available perfect public servant never materialised. What is clear is that at that time, adapting a virtual agent to a general purpose local authority was no easy task.Ten years on, and I am about to publish an updated version of the book amid a whirlwind of hype and hope for AI. We find local authorities are increasingly experimenting and adopting different forms of AI – for the same four reasons: cashable savings, meeting expectations, improved compliance and offering a more personalised and compassionate service.Much of the adoption for now happens in the back office, with AI summaries of Teams meetings being the obvious example. There is less confidence in unleashing AI into frontline work.Perhaps the most notable area of growth is in AI transcription tools, particularly in frontline social work. This is an example of AI augmenting frontline work rather than directly substituting it.Ten years on, Amelia might not be on the scene, but visit any local authority website and you will be met by a chorus of chatbots: Ali, CeCe, Helen and SAM; ELLIS, Tom, Ker-Know and IMPY; AiDA, Bailey, B-Bot and BOB; Bobbie, Nelly and SARA and Aiimee; WatBot, Ami, Gracie, Matti or maybe Nia, Brum Bot and, of course, Darcie.The primary role of these chatbots is to signpost users to the right information which, done well, can often deflect demand on the call centre.Derby City Council's Darcie has attracted the most attention as it operates across different channels in 10 languages and has handled more than 3.5million enquiries since its launch in 2023.What these chatbots are not, is animated photo-realistic emotionally-embodied avatars – except for Jimmy at Visit Liverpool, perhaps.The boom in generative and agentic AI over recent years is about ever more advanced and autonomous agents handling tasks – but it is not about embodying as a fantasy figure, the perfect public servant.Amelia continued to be deployed in HR and finance settings but has since migrated to being a platform rather than an avatar.The platform was sold to new owners in 2024 and is now supporting voice calls in the US private health sector, where you can find demonstrations of it flexing its agentic muscles – showing how it can process a prescription, raise and process an invoice and book some medical tests all in one phone call – and all without a human in the loop.The perfect virtual public servant never arrived, but perhaps Amelia's story leaves us with a more interesting question: what exactly was this perfection we were striving for?Ten years on, human public servants remain on the frontline, increasingly augmented by and learning to live and work alongside the very technologies that once seemed destined to replace them.Stephen Jeffares is an associate professor in public policy at the University of Birmingham's Institute for Local Government Studies (INLOGOV), specialising in digital government, public sector AI and social media data analysisThe Virtual Public Servant, AI and Frontline Work 2nd Edition 2026 will be published on 8 October by Palgrave Macmillan

Title

AI

Rethinking the public servant for the AI era

By Stephen Jeffery | 07 August 2026

Popular articles by Stephen Jeffery

SUBSCRIBE TO CONTINUE READING

Popular articles by Stephen Jeffery