Call Summarisation – What is it and who wants it?

Finding Calls When You Need Them

When Threads was first conceived back in 2012, transcribing calls was not that great. People would often look at call transcriptions and laugh at some of the mistakes it made. However, what we were at pains to explain is that the major benefit of transcribing calls is not in providing a perfect written record of the call, but that you can find the call when you need it.

This is easy enough 10 mins after making the call, but a week, a month, a year later, it is much more difficult. If you know who you spoke to and exactly the date and time of the call, then it is straightforward enough, but who remembers that?

Indeed, it is exactly the same when finding emails. Why should it be any different finding phone calls? They are just as important.

The Nuances of Speech Recognition

What you find with speech recognition – which is the technical name for the transcription process – is that it is much better at recognising complicated words  like “refurbishment” than it is at recognising simple words like “and” or “the”. And when you think about it, if you searched all your emails for those containing “the”, you would probably end up getting most of them. But not many containing “refurbishment”.

Beyond Transcription: Efficient Navigation

So a key benefit of call transcription is that it enables users to search for relevant calls, and to do that, the speech recognition doesn’t actually have to be that good.

But what do you do when you find the call you are interested in? Speech recognition is much better now than it was 10 years ago, but who wants to wade through even a perfect transcription of a 30 minute call, trying to find exactly the point where someone says, “yes, it was my fault”.

So Threads had the unique ability to listen to the call while watching the transcription and scrubbing through sections of the call not of interest. You can see a demonstration in the video below. What you notice is that not only can you scrub through a call to find the part you want, but also listening to the call while reading the transcription makes the transcription errors hardly noticeable.

AI-Powered Call Summarisation

But now that AI has advanced so much, we can go a step further summarising the transcription, so you don’t, in many cases, need to read it at all. Have a look at this short video and see  how the real-life 11 minute call is found, listened to and summarised. Enough said?

This still takes our breath away. Why not contact us to discuss how you can transcribe and summarise your calls with Threads. Or look at an existing Threads case study.