AI dictation tools are changing the future of office work

Can you imagine walking into an open office on a weekday, seeing people at their desks and hearing practically nothing? Not even the staccato rhythm of fingers on a keyboard?

It certainly seems that this could be the future of work, in all types of professions. Well, at least the ones that involve people working in front of computers.

You’ve no doubt heard a lot about AI disrupting productivity in recent years, with the technology promising to speed up workflows and advance progress in numerous research-intensive fields.

It turns out that the literal way we work, that is, what we do while sitting at our desks, is also transforming.

AI increasingly allows people to ditch their keyboards and simply dictate what they need to do. That includes everything from sending an email to writing a presentation to coding custom applications.

Dictation tools have been around for decades and have largely become commoditized to the point that the on-screen keyboard app on your iPhone or Android device comes with a free one. In recent years, they’ve gotten especially good, thanks to the emergence of widely available machine learning models for speech recognition and many companies building apps on top of it.

What has happened recently is that there are now a wide range of speech-to-text apps that are surprisingly good at their job, to the point that you can realistically use them at work and achieve a high level of accuracy and speed with your dictation.

Today, you can have an open source dictation app like Handy run a completely local machine learning model on your computer and ensure your speech is never sent to the cloud for processing, keeping your data private. You can also opt for an app like Wispr Flow, which uses larger models to clean up your ums, ahs, and mispronounced words into text that requires less manual editing.

Handy is the app I’ve been using for the past few months, by the way. I love that it’s easy to activate with a keyboard shortcut and that it provides unlimited use at no charge. Wispr Flow, on the other hand, is currently targeting the enterprise market, meaning it wants to sell its dictation and note-taking service to businesses. The company behind it just raised a $280 million funding round, at a $2 billion valuation.

There are also interesting things happening on the hardware side of dictation. Since January, Subtle, a San Francisco-based startup, has been selling its $250 VoiceBuds that look like regular wireless headphones, but work with the company’s cloud-based service to pick up your speech, even if you’re in a crowded space talking to people or whispering softly so as not to be overheard.

Sutil says that, interestingly, its headphones don’t rely on high-end microphones to hear you. That’s all real-time AI working in the background to distinguish what you’re saying from background noise. The company claims that its Voicebuds produced up to five times fewer transcription errors than Apple’s AirPods Pro 3 combined with OpenAI’s transcription model in noisy environments.

I suspect that Subtle will face competition in the very near future, meaning more people will likely want to get things done at work using headphones instead of a keyboard. This is not just because some people don’t like or are not good at writing, but because the scope of what can be achieved by simply issuing a command is expanding rapidly.

In other words, voice is becoming an increasingly powerful input mode for computing. The basic hardware to use it is already widely available, and services that improve your input by clarifying your speech, inferring what you mean in specific contexts, and taking actions are evolving as we speak.

It may seem strange right now, but whispering sweet nothings into your device at the office could soon catch on.