Agent tutorial, Clarabridge integration, Anonymizer. And onboarded new customers in energy, services, telco including our first Belgium customer! 👋 🇧🇪
The new tutorial helps onboarding agents, learning them everything they can do with Deepdesk.
In 10 minutes all features are introduced, with tips and tricks on how to work smart and fast and get the most out of Deepdesk.
We implemented the capability to anonymize all the message data that we process. Our custom developed named entity recognizer works real time on all in- and outgoing messages of agents and customers as well as training data and analytics.
For one major account we have a custom GPT-2 model live for one some time now. Agents are happy and occasionally surprised with the suggestions. The model is good for around 10%-15% of all assisted text. We’re currently monitoring the model output, and looking into improving overall quality, lowering inference and auto scaling to reduce costs. A second model is being tested at another customer with a few agents.
We’ve updated our training algorithm to improve personalization. Messages that agents use more often have more chance of showing up for that agent. A/B tests showed an increase in usage of the new model, and after switching completely overall scores increased.