Transforming how governments serve citizens through voice AI

As AI systems develop, the limiting factor will not be the quality of the technology itself, but how users interact with them.

This was the core message of the presentation delivered by Hugo Rayne, GTM Lead at ElevenLabs, at Government Transformation Summit on 25 June.

One logical solution to this problem, Rayne argued, is for public services modernising with AI to adopt the mode of interaction which comes most intuitively: speech.

AI voice technology is exactly what ElevenLabs is specialised in, with it offering both conversational agents and creative tools. Its conversational agents, which are of greater use for government, consist of text-to-speech and speech-to-text systems layered with a Large Language Model, allowing users to simulate authentic interactions with an advisor or expert.

Such models could prove useful in government contact services, permitting citizens to receive relevant information or guidance immediately, at any time of day, and in a variety of languages, Rayne said. This would allow callers to avoid the potentially frustrating experience of waiting in queues or navigating phone menus.

Human input from the service provider’s side, however, remains fundamental, and Rayne stressed how the company's models could direct more complex cases to the appropriate person once relevant context had been captured, ensuring that humans are not replaced but instead that access to services is simplified.

ElevenLabs agents are already in use in a number of contexts. Notably, the company created a voice clone of Mykhailo Fedorov, the then Ukrainian Deputy Prime Minister and Minister of Digital Transformation, to answer questions about policy. Further integration of voice AI into Ukrainian public services is planned, as the country seeks to make public services more accessible.

Further examples of its application can be seen both at the local authority level and in the delivery of social services. The city of Midland in Texas created an outflow agent called Jackie from ElevenLabs Agents, leading to 7000 fewer missed calls across five languages. The UK social services enterprise Beam also adopted ElevenAgents for out-of-hours phone line responses, transcription, and translation, observing a 50%+ reduction in call handling time.

In addition, ElevenLabs technology led to a 25% onboarding reactivitation rate for Deliveroo drivers who had dropped off the onboarding reactivation programme, and is being used to help those who have lost their voice due to degenerative illnesses.

 However, technical performance and user experience alone are not enough to make AI suitable for government use, Rayne argued. Before ending the talk by re-emphasising the potential of voice AI to reduce bottlenecks, he highlighted the importance of regulation in AI systems. 

He said: “The quality of the model and product gets you in the door, but to get you over the line, you have to be able to show that you are extensively covering all angles, from data privacy, security, compliance.”