There is an old, apocryphal saying that people listen to 20% of what you say and 80% of how you say it. While that may have frustrated plenty of subject-matter experts over the years, it reveals an inconvenient truth: tone, cadence, and vocal delivery often matter more than the content of your speech.
A new generation of technologies is now taking this a step further. Vocal intelligence tools, such as empath_AI, use 'vocal biomarkers' – measurable acoustic patterns in the human voice – to spot when someone may be under strain, struggling to follow, or in need of additional support. It is a way of looking beyond people's words, helping frontline service providers in the public sector adapt their conversations to the situation of the person they are speaking to.
The principle is already embedded in regulation like the Financial Conduct Authority's Consumer Duty, which requires firms to put customers' interests first and treat those in vulnerable circumstances fairly. In the public sector the equivalent obligation is statutory, under the Equality Act 2010 and the Public Sector Equality Duty. But all of these frameworks share a major flaw: they assume people are willing to share their situation with a stranger, or even realise they are struggling at all in the first place.
Anyone who has been through a difficult period will tell you the effects are physical as well as emotional. Your emotional state triggers changes in the nervous system that ripple through the body, including the muscles and breathing patterns that control the voice. The voice responds to stress, fatigue, or cognitive strain, creating subtle acoustic variations that are often detectable before a person consciously realises they are struggling.
These changes are far from random – different combinations form recognisable patterns. Without realising it, you will have done it yourself. When you have come home from work and asked your partner or flatmate how their day was and they tell you "it was fine", but you know from their tone it was not, that is your brain picking up on how the words were said. Vocal biomarker tools formalise this intuition, analysing the underlying acoustic features that sit beneath language itself.
The voice carries a wide range of variation across pitch, cadence, vocal energy, breath flow, and micro-patterns that can be measured in milliseconds. Someone under acute strain may sound tighter or shakier as throat muscles tense and airflow restricts. Low energy states tend to produce quieter, flatter speech. Where processing has become effortful, the pauses between words lengthen – a kind of 'cognitive stutter' as retrieval slows.
One way of thinking about these technologies is as hearing what a well-trained ear would catch on a good day, consistently, on every conversation, including the shifts too small to notice in the moment. It does not diagnose conditions or apply clinical labels and the person on the other end of the line does not need to say anything about their circumstances. Instead, the tool alerts the member of staff that the person they are speaking to may need a gentler pace, clearer explanations, or additional support in a conversation.
These systems use advanced machine-learning models trained on large datasets of vocal recordings, with roots in clinical research. Rather than focusing on what people say, the models analyse how they say it: the variations in pitch, breathiness, timing, and vocal-fold vibration that shift when someone is under strain. They identify patterns that are statistically associated with moments when people may need additional support.
Using vocal biomarkers originated in clinical research, but the technology is now being applied in adjacent sectors – and adoption is already underway in public services in other countries. In Peru, for example, the National Police is exploring it as part of its duty to identify and support officers exposed to occupational trauma, set out in a 2020 police health directive.
The potential applications across UK public services are extensive. Whether it is policing, fire, or ambulance services, earlier recognition of strain could help prevent crises, improve decision-making, and support frontline workers who routinely face traumatic situations. Where services already offer confidential support lines, it could help them recognise who needs a bit more, so support is offered earlier and by consent rather than after a crisis.
In prisons, for example, staff could be alerted when a prisoner sounds distressed on a call or at reception, supporting the judgement they already exercise under existing safety procedures rather than replacing it. The Ministry of Justice is already exploring technologies that support staff wellbeing and reduce burnout – a growing issue as services feel the strain of tight budgets and recruitment and retention challenges.
But more broadly, it also aligns with wider efforts to keep people in the workforce. Public services are among the country's largest employers, and initiatives like Keep Britain Working emphasise the importance of supporting individuals before they fall out of work and struggle to return. Earlier recognition of strain could help public services provide timely assistance, prevent burnout, and improve retention – which is absolutely critical at a time when finding the right talent is proving difficult.
That wider context is absolutely critical. Pressure on the country's mental health remains historically high, particularly among younger generations – according to NHS England, around one in five people between the ages of eight and 25 had a probable mental health condition in 2023, a rate that rose sharply from 2017 and has stayed there since. Public services need to work in that context and prevention is consistently more effective – and more cost-efficient – than cure.
Ultimately, whatever their circumstances, people want to be treated with dignity. Vocal biomarker tools do not replace human judgement or compassion – if anything, they can enhance them. By giving frontline professionals better insight into how someone is managing the conversation, it helps them respond with empathy, patience, and care.
Public services are at their best when they recognise the humanity of the people they serve. Vocal biomarker technology offers a way to strengthen that connection, ensuring that the state listens not just to what people say, but how they are saying it.