In today's fast-paced business landscape, delivering exceptional customer service is essential for success. AI-powered smart routing, such as Sprinklr’s Smart Pairing, is a game-changer in the world of customer support. This feature leverages customer data and AI algorithms to intelligently match unassigned cases with the most suitable agents, resulting in quicker query resolutions and improved care Key Performance Indicators (KPIs). In this article, we'll walk you through the process of setting up AI-based smart routing for your contact center.

Important Pre-requisites

Before we start implementing smart routing, it's important to ensure that we have the necessary pre-requisites in place which are discussed below :

• Issue Intents: To get started, we identify the different types of issues or inquiries your customers commonly have. These could include issues like "Repair," "Refund," "Technical Support," and more. Each issue intent should represent a specific category of customer inquiries that require expertise to handle effectively.

• Coverage and Case Diversity: For increasing the effectiveness of Smart Routing, we have to ensure that there is a wide coverage of inbound cases in the platform that correspond to the identified issue intents. This diversity of cases helps the algorithm learn and make informed decisions when matching cases to agents.

• Sufficient Case Data: For each issue intent, we gather a substantial number of historical cases. It's recommended to have at least 100 cases for each intent. These cases will serve as the foundation for training the smart routing algorithm and improving its accuracy over time.

• Survey Scores and Agent Feedback: Next, we collect customer survey scores such as Net Promoter Score (NPS) and Customer Satisfaction (CSAT) ratings. These scores provide valuable insights into customer satisfaction and the performance of agents. Additionally, any feedback or scores related to agent performance can be used like predicted CSAT, or partner-specific metrics that can help assess the agent's capabilities.

• Agent Pool Availability: Apart from covering cases, we should also have a sizable pool of agents available in the queue where smart routing will be implemented. Having a diverse group of agents allows the algorithm to make informed decisions and optimize case assignments based on agent expertise and availability.

Steps to Set Up Smart Routing

After we ensure that the above pre-requisites are met for a queue, we do the the following steps internally to set up the process of routing.

Step 1: Data Collection and Calculation

• Calculate Weighted Average Scores: Calculating weighted average scores for agents across individual intents and at an aggregate level using historical case data. This involves assessing agent proficiency in solving issues, perceived service quality, and other relevant criteria.

• Automatic Training Job: Setting up an automatic training job to continuously update and refine the agent scores. You can choose the frequency of training updates, such as hourly, daily, or weekly, depending on your needs.

Step 2: Agent Ranking

• Agent Ranking: Assigning a rank to each agent based on the above calculated scores. The lower the rank, the better the agent's suitability for handling cases.

• Available Agent Selection: When an incoming case needs to be routed via smart routing, the algorithm selects the best-ranking (lowest rank) agent who is also available at the time of assignment.

Step 3: Fine-Tuning and Customization

• Configuration Options: We explore various configuration options to fine-tune the scoring algorithm according to your specific business requirements. For instance, we can set thresholds for the number of cases an agent should handle before being considered for smart routing.

Step 4: Impact Measurement

• Visualize Impact: Lastly, we use care reporting module to visualize the impact of enabling smart routing. For example, we can compare survey scores before and after implementing smart routing to assess the effect of smart routing on customer satisfaction.