Voice is one of the hottest, buzziest and most over-hyped technologies in an arena already stuffed with over-hyped technologies. We all want to Alexa power our fridge, although none of us is exactly sure why. In fact, music streaming and getting news and weather are the most popular uses for home voice assistants. To paraphrase The Hitchhiker’s Guide to the Galaxy’s resident cynic, Marvin the Paranoid Android, “Brain the size of a planet, and all they do is ask me to turn on the radio…”
In the Enterprise, everyone is racing to voice enable their business processes, and to have a cool Alexa skill. But for most companies, using a public cloud provider puts them at a security and regulatory risk. There is no “magic pipe” between Alexa’s microphone and your data centre. Everything a consumer says is transcribed and analysed by the cloud provider, and everything you send back goes down the same path. If you are a bank, should your customer really have to have their financial details exposed this way? Should personal medical details be available to cloud providers? Until there is a secure way of connecting voice assistants to the provider of information, in the same way we look for the SSL padlock on a browser, there are so many use cases that voice assistants in their current form just cannot fulfil.
Fortunately, we are beginning to see the growth of edge–based devices that can perform basic transcription, even intent analysis, and the tools exist in any event for you to take audio from your customers’ browsers and phones and process them securely in your own environment.
But actually, this fluffy front–end interaction may well be the least of your problems, depending on the type of data you already process (don’t worry, there is some good news at the end). There is a pretty good chance that someone somewhere in your organisation is generating voice content: It could be in a call centre; it could be videos from marketing or maybe recordings of Skype and Zoom calls. As regulations creep up, first in Europe with GDPR and now in the US with CCPA, the detail you need to know about this content is vastly increasing.
If I put in a Subject Access Request to your organisation today and said I called you from an unlisted number a few weeks ago, could you find that call without having someone listen in to it all? Or I asked you to delete any phone calls I was on; how could you guarantee that if I had made calls from my home number. What if my wife had called from that number as well? Would you end up deleting all her calls? The fact is that metadata on phone calls is far poorer than that available for, say, emails, and that means you are probably doing a poor job of reconciling your audio files with the available customer data.
Run IT for a bank? You are probably required by your regulator to put in place robust policies against Market Abuse and Anti-Money Laundering, with all sorts of retention requirements under MiFID II. To achieve that, you need to be listening to every single phone call made by every single trader, and that is a lot of phone calls. And you are probably having to store them for 5 years as well. And when the regulator turns up, you need to be able to find them in a hurry.
Unfortunately, most CIOs see voice as a challenge. How do I overcome privacy concerns? How do I get value out of voice data? Isn’t speech recognition of phone calls inaccurate? What if people switch languages to evade my surveillance? How do I find the voice needle in the haystack?
Fortunately, lots of industries are getting value out of voice.
In financial services, banks are putting in place robust communications monitoring, running in a fully secure and private environment covering hundreds even thousands of traders.
Government departments are performing automated QA of their phone calls to increase citizen engagement and satisfaction, allowing sample sizes to rise from less than .5% to cover every singlecall.
Litigation is now faster and cheaper, with 70% increases in audio review times (which is good news for everyone, except lawyers).
The Police are finding ways of better handling non-emergency calls to free up operators to concentrate on priority emergency calls.
Even motor sports is getting in on the action, monitoring driver and pit radio traffic in real time to gain that tiny bit of competitive advantage.
Whichever industry you are in, you can find a way of leveraging voice data in a secure, easy-to-deploy and scalable way, without the worry that your precious data is being listened to in an office block on the outskirts of Bucharest.
