rewrite this title Your AI Tools Are Only as Good as Your Audio – UC Today

The promise of AI in the workplace is compelling.  

Meeting summaries, action capture, intelligent search, copilots that surface what was said and what was agreed – the tools are here, and adoption is accelerating.  

But for many organisations, those tools are quietly underperforming. Not because the technology isn’t good enough. Because the inputs aren’t. 

When AI has to work from meetings plagued by background noise, crosstalk, inconsistent microphone pickup, or broken audio, the outputs suffer.  

Summaries miss nuance. Action items are incomplete and transcripts are unreliable.  

The AI is only as good as what it hears. 

This is the conversation Richard Trestain, Product Marketing Manager at Jabra, thinks the industry needs to have more openly.  

“We’re all talking about AI as a software conversation,” he says.  

“But the quality of what AI captures depends entirely on the quality of the audio going in. You can’t separate the two.” 

The Noisy Reality of Hybrid Work 

Hybrid work has matured. Offices are filling back up, but the flexibility is permanent – knowledge workers routinely move between offices, home environments, and third locations throughout their working week.  

Every one of those environments has noise. Open-plan offices are louder, partly because collaboration is more constant.  

The person next to you is on a call, or you are on a call. And increasingly, people are talking to AI assistants out loud, adding yet another layer of ambient sound to the mix. 

The result is that the background conditions for any given meeting are unpredictable and often poor.  

“Everyone has got noisier,” Trestain says. “The office has come back, but most people are still working from multiple locations. It’s all about flexibility – and most are noisy environments.” 

For AI tools that depend on clean audio capture – real-time transcription, meeting summaries, voice search – this is a structural problem.  

Not a one-off, but a consistent pattern that scales across every call, every day. 

From Sympathy to Expectation 

There’s a cultural shift happening alongside the technological one. When hybrid working was first taking off years ago, people were forgiving of poor audio, but that grace period has passed. 

“The sympathy has worn off,” Trestain says. “The second someone’s voice breaks up or there’s big noise in the background, people say very quickly: can’t hear you.” 

He draws a useful analogy: using a professional headset is the audio equivalent of a virtual background. “We’ve all learned to put our backgrounds on in video calls. A headset is kind of the equivalent of a background blur for sound. The second any noise creeps in, it kills the meeting.” 

What this means in practice is that poor audio is no longer just a comfort issue – it’s a trust and credibility issue. And ultimately, this has a significant impact on productivity. “If you’re the one trying to make people change their minds, or it’s your meeting to run, you’ve already put yourself at a massive disadvantage by not sounding good.” 

“It leads to an erosion of the way people think of you.” 

That erosion of trust extends to AI outputs too. When your meeting summary is partial, or the action items from a call are wrong, confidence in the tooling drops – even when the real culprit was the audio going in. 

Noise Cancellation: More Than a Feature 

One of the most misunderstood aspects of professional audio is what noise cancellation actually needs to do – and there are two distinct aspects to this. 

The first is outbound: ensuring that what you say is transmitted clearly, with background noise removed from the microphone feed before it reaches other participants or an AI transcription engine.  

The second is inbound: protecting the listener from their own environment so they can focus and follow the conversation without their brain working overtime to filter out distractions. 

Trestain explains that most enterprise headsets don’t provide active noise cancellation (ANC) during calls.  

The technical challenge is sidetone – the natural feedback of your own voice into your ears that makes speech feel natural – and doing that simultaneously while also blocking ambient sound is genuinely difficult. 

“What we’ve managed with the Jabra Evolve3 75 and 85 is to do both of those things at the same time. We’re the only headsets to have ANC during calls, with sidetone. You’re protected from the sound around you, you hear your own voice naturally, and people hear you clearly. That’s the utopia.” 

For AI meeting tools, this matters more than organisations typically realise.  

Every layer of noise in the audio feed degrades transcription accuracy and therefore the quality of every downstream output: the summary, the action list, the searchable record.  

Clean capture at the input stage compounds positively through the entire AI workflow. 

The IT/UC Standardisation Case 

What prevents most organisations from solving this isn’t just awareness – it’s standards and personal convenience. The most common pattern Trestain sees is a permissive approach to personal devices: employees bring their own headsets, often consumer earbuds worn on the commute, and use them for professional calls without a second thought. 

“Consumer devices aren’t designed to speak perfectly to the platform. They’re not designed to be managed by IT at scale. They’re not particularly great at cancelling noise from the call. So generally, you get a sub-par experience.” 

The organisations that consistently achieve better outcomes, he says, are those with clear device standards. “The greatest success occurs with organisations who are rigorous with their implementation of professional audio devices – providing employees with headsets that they are happy to use everywhere, not just in the office .” 

An AI-Readiness Baseline for Audio 

If organisations want AI collaboration tools to deliver on their promise, the starting point is clear: treat audio quality as infrastructure, not a personal preference. 

That means defining role-based standards – contact centre agents, office-based knowledge workers, and road warriors have different needs, but all need a defined minimum.  

It means choosing devices designed for enterprise environments: platform-certified, remotely manageable, purpose-built for call quality in unpredictable conditions.  

And it means thinking about the full picture: where are your people working? What are the noise conditions? Are their devices doing two jobs well – personal listening and professional communication? 

“If you pare it all down,” Trestain says, “it’s essentially comfort and call quality. But when you layer AI on top of that, the inputs matter enormously.” 

Read more: The 14% Signal: Why You Need to Prepare for the Voice AI Surge