What’s next for ambient scribes? Healthcare's chaos zones

Ambient AI is fast becoming the breakout use case in healthcare technology, pushing aside skepticism in a largely cautious sector.

At a recent major healthcare conference, we spoke to every company working in ambient scribe. The pattern was consistent: the conversation has shifted from "if" to "where next."

The technology has proven itself in the controlled settings: the GP surgery, the outpatient department. One patient, one clinician, a door that closes. Of all the things AI can do in healthcare - interpret medical images, provide clinical decision support, predict patient outcomes - ambient scribe might seem like the modest option. 

But it's winning because it solves a problem clinicians actually want solved: the documentation burden that steals time from patient care.

Doctors already ranked their priorities

If you ask clinicians where they actually want AI, the answer is revealing. In a recent survey of 2,000 clinicians by Lucy Goodchild, two use cases sat right at the top: writing patient letters and clinical notes, and analyzing medical images.

Everything else was further below...

Adrian Mulligan, Chris West, Lucy Goodchild, Nicola Mansell | Clinician of the Future 2025

Those two matter because they map directly to what speech technology can already do well.

Voice in, structured text out.

Before we get carried away with science-fiction agents, the most popular real use cases are incredibly practical.

Help me write what I have to write. Help me keep track of what I'm seeing.

Ambient scribes are already earning trust in that narrow band of tasks. The question now is whether it can follow healthcare into its chaos zones.

Into the chaos zones

Take the example of a resuscitation bay in a major emergency department. You've got a doctor and two or three nurses surrounding a patient. There's a lot of chat. A lot of critical decisions are happening fast.

"Give the patient more blood." "Do this, do that."

If that conversation was being recorded and structured into data, it could provide a time-stamped record of who said what and when. Who gave the instruction for more blood, and at what time. When the decision to intubate was made. Which nurse confirmed the dose.

That's not just documentation. That's an audit trail for the most critical moments in patient care.

The same opportunity exists in the back of an ambulance. These are tougher environments than the consulting room, places where ambient scribe is only just starting to be explored.

There are also other environments that aren't chaotic, just complex, such as endoscopy. It would be possible to have an ambient scribe running while the endoscopist is doing the examination. They could talk through what they're seeing and then that could be summarized. 

The challenge is the same across all these settings: taking technology that works in quiet rooms and making it work where medicine actually happens. And that means solving some hard technical problems.

What stands in the way

You need super high accuracy in noisy environments. Not marketing-deck accuracy on clean audio, but stubborn accuracy when there's an alarm in the background, someone coughing, someone crying and three clinicians speaking over one another.

You also need to handle dialects. Healthcare is one of the most linguistically diverse environments you can find.

A London emergency department might see accents from Glasgow, Lagos and Gdańsk in a single morning. If your scribe falls apart every time it hears a non-standard accent, it's unusable.

The subject of data residency comes up a lot in discussions we're having at the moment. In some regions, regulators and hospitals simply won't accept patient audio leaving the country. If you want to work in those markets, you need on-premise deployment options. Medical models need to work seamlessly across both SaaS and on-premise environments without compromising accuracy.

Get these things right and the opportunity opens up. Get them wrong and ambient stays in the consulting room.

The reason to solve them isn't just about expanding to noisier environments. It's because looking to the near future, voice won't work alone. The future isn't ambient scribe as a standalone tool. It's ambient scribe as the interface layer between clinicians and multiple AI systems working together. 

From consulting room to clinical nervous system

The quiet consulting room proved the concept. It allowed the technology to mature and clinicians to build trust. But the back of an ambulance, A&E, resuscitation rooms - these are healthcare's chaos zones, and they're where the technology needs to go next.

These are the places where documentation burden is highest and where accurate, automated capture of clinical dialogue could make the biggest difference.

Not replacing clinicians or making clinical decisions, but being present in the moments that matter most, capturing what needs to be captured so clinicians can focus on what actually saves lives.

If we can nail accuracy in noisy environments, handle diverse accents and dialects, and manage complex governance requirements like data residency - areas where Speechmatics has focused its strengths and development - ambient scribes will stop being a clever transcription tool and start becoming part of the clinical nervous system.

The chaos zones are waiting.