By Alejandro Heredia on Nov 16, 2020 11:13:59 AM
Welcome to the end of 2020. We’ve almost survived it! We are on the path to digital, interconnected. We have gone from multichannel to omnichannel; anytime, anywhere. Like us people; businesses and enterprises have followed and the customer is there to join them.
Now, it’s no secret we have been recording everything. We star our favorite audios on the chat, we save our favorite conversations. Companies have larger and better repositories of data for everything we do. And yes, we have been recording our customer’s interactions for a while now.
But, what do we do with all that information? We mine it, we analyze it, we report on it so, What else can we do with it?
We have millions of audio and interactions with our customers. Our Speech-to-Text Technologies and Automatic Speech Recognition engines are gathering everything and we have agents on the other side but we also have Independent Voice Recognition tools, Chatbots, Web-bots attached to AI and Machine Learning brains. It is time to begin to answer back correctly to our customers.
Which brings me to the issue at hand, the issue with PII: Personal Identifiable Information.
We have all the data, does it mean I can use it? Can I grab all the audio data and start really analyzing what my customers are saying and continue gathering, labeling, feeding them to my Machine Learning model so that I can talk to customers humanly and understand what they are saying?
Most likely, depending on your industry, the answer is: I can only use some part of it or even I’m totally blocked because of Personal Identifiable Information contained in such recordings. A lot of exchanges with customers are directly classified, think of Telemedicine or credit information exchanges with banks.
So, what’s the workaround? How do I jump this hurdle if I need data, I need language-specific data, domain-specific, intent, and prompt specific data to make sure my responses are more human?
To be able to understand and talk to your customer, you will need to constantly train your machine while protecting customers' data ownership and privacy.
That’s one of the issues we at Atexto came to solve. We want to help you improve your Recording and Transcribing capabilities by benchmarking with the best ASR engines. But we went further and we can collect and label the data for you to model your Machine Learning capabilities without compromise of PII. We can work with your data, and we can provide new Independent and Identically Distributed datasets that mimic your customer’s reality or your customers’ customers.
Originally published as an LinkedIn Article here.