[Music] let's have a look at admin console an exciting feature in task intelligence for Tokyo today customers see the power of AI to solve previously unsolvable business problems but while AI models offer promising outcomes there's much anxiety regarding adopting such solutions for example customers may have anxiety over allowing a model to set assignment groups for a case or setting a case as priority they may also question accuracy if a model's recommendations differ from that of human entered data admins may not know the nature of data being processed and may want to place control in the hands of business users business users on the other hand need a simple approach for setting up and adopting models let's see how the admin console helps address these challenges in filter Navigator let's start by typing task intelligence for customer service and selecting case setup here is the home page for the admin console the home page is your One-Stop shop to rapidly configure train and deploy AI Solutions on the CSM platform here on the home page we can choose an active model and in real time see statistics associated with that model such as the number of cases with predictions and the number of predictions made in the last seven days by clicking view all metrics we can see the business value of the model and the predictions used by agents over time going back to the home page we can see in the middle the models that have been currently deployed and at the bottom of the home page this is where we've taken a different approach we have moved to a use case driven experience versus technical AI jargon so for example in this first use case it's about predicting case values and choices to reduce handle time in the Second Use case it's about predicting case sentiment to improve customer satisfaction and coming soon there are a third use case to predict case language to improve assignment logic let's have a look at a model that has been previously deployed clicking on the last Model here we can see that there is a simple five-step process to setting up a model in Step number one you define the purpose of the model will this model look at inbound emails or inbound cases if it looks at emails or cases will it also interrogate the associated attachments in the second step we train our model here we Define the output table and output Fields the input table and the input fields and perhaps most importantly the conditional logic that will determine which records the model will be trained against it's important to note that configuring this isn't done in a low code no code Banner while within the reach of a business user the next step is to assess the model in this example the model will autofill up to 1052 field values more importantly we can see on a field by field basis where the model would make predictions that match that of an agent that are different from an agent or perhaps where the model has skipped over records by clicking on view sample results we can see on a record by record case-by-case basis where the model's predictions occurred let's go back to setting up the model the next step is to set your preferences in this step we're able to determine which field values the model will predict and for those field values whether it will make a recommendation or whether it will perform an autofill so it's important to note here that business users have very fine and granular control over which Fields will be predicted and how those predictions will manifest themselves on the UI this addresses a key deployment anxiety customers can adopt predictions at their own pace on a field by field basis for a very conservative customers they can also run models in the background no recommendations or autofills will appear on the UI but they can still see predictions at the table level these are options that customers can turn on or off based on their own risk tolerance in Step number five we have a review of our model we have a review of the field of the records that the model has been trained on and what the model will do with respect to which Fields it will predict and which Fields it would make recommendations on and in one simple button we can deploy the model so there we have it a simple five-step approach to set up train test and deploy AI models our approach helps reduce anxiety related to adopting AI in the real world customers can start with recommendations and move to auto field on a field-by-field basis they can gain insight into the business impact of predictions versus Agent entered data and perhaps most importantly they can put the power of AI in the hands of business users the people with a pulse on the business thank you