• Written by Sonali Subhadarsini, Product Manager

A customer service manager at a large retailer is deciding this week how to handle a surge in complaints about a new product, and currently they base this decision on a manual review of recent tickets and their own experience with similar issues in the past. However, this approach can be time-consuming and may not take into account all the relevant information, such as the product's design and manufacturing process, or the customer's previous interactions with the company.

Characteristics of agentic workflows

An agentic workflow for customer service is distinguished from a simple chatbot by its ability to use various tools, maintain state over time, and work towards a specific goal, such as resolving a customer's issue or improving their overall experience. This requires a more sophisticated system that can integrate with multiple data sources and systems, and make decisions based on that information. For instance, a customer service agent may use a knowledge base to find answers to common questions, and then escalate the issue to a specialist if necessary.

The shift supervisor at a call centre, for example, checks the maintenance log to see if there are any known issues with the product that could be causing the complaints, and then uses this information to inform their decision about how to handle the surge in complaints. This kind of tool use and state maintenance is a key characteristic of agentic workflows, and allows them to provide more effective and efficient customer service.

Human approval and oversight

Most people assume that removing human approval steps is a key benefit of implementing agentic workflows for customer service, but this is often premature. While it is true that agentic workflows can automate many routine tasks and decisions, there are still many situations where human oversight and approval are necessary, such as in complex or sensitive cases. The category manager at a retailer, for example, may need to review and approve any changes to the product's design or manufacturing process before they are implemented.

In fact, removing human approval steps without proper consideration can lead to errors or unintended consequences, such as a failure to properly escalate issues or a lack of transparency in decision-making. A more effective approach is to identify areas where human approval is still necessary, and design the agentic workflow to include these steps in a way that is efficient and effective.

Dependent data systems

Agentic workflows for customer service depend on a range of data systems, including ticket history, knowledge base, product and order records, and escalation history. The customer service representative, for example, may need to access the customer's previous tickets and interactions with the company in order to understand the context of their current issue, and then use this information to inform their decision about how to handle the issue. This requires a high degree of integration and interoperability between different systems and data sources.

For instance, the product and order records may be stored in an enterprise resource planning (ERP) system, while the ticket history and knowledge base may be stored in a separate customer relationship management (CRM) system. The agentic workflow must be able to access and integrate this information in real-time, in order to provide effective and efficient customer service.

Impact on customer service processes

Agentic workflows for customer service can have a significant impact on customer service processes, particularly in terms of reducing repetition and improving root cause analysis. Currently, many customer service issues are answered from scratch each time they arise, without any consideration of the underlying causes or previous interactions with the customer. This can lead to a lack of consistency and effectiveness in customer service, as well as a failure to identify and address root causes of issues.

By using agentic workflows, customer service teams can identify and address these root causes, and provide more effective and efficient service to customers. For example, the customer service team may use an agentic workflow to identify a common issue with a product, and then work with the product design team to implement a fix. This can help to reduce the number of repeat contacts and improve customer satisfaction.

Evaluation metrics

Agentic workflows for customer service can be evaluated using a range of metrics, including first contact resolution, time to answer, and repeat contact rate. These metrics provide a way to measure the effectiveness and efficiency of customer service, and identify areas for improvement. For instance, a company may use these metrics to compare the performance of different customer service channels, such as phone, email, or chat, and identify opportunities to improve the customer experience.

In order to get the most out of these metrics, it is necessary to have a clear understanding of what they measure and how they can be used to improve customer service. This may involve working with a data analyst or other expert to design and implement a metrics and evaluation program, and using this information to inform decision-making and drive improvement. Customers can contact us to learn more about how to implement agentic workflows and evaluate their effectiveness.

Limitations and exceptions

While agentic workflows for customer service can be highly effective in many situations, there are also limitations and exceptions to consider. For example, agentic workflows may not be suitable for highly complex or sensitive issues, such as those involving emotional or psychological support. In these cases, human customer service agents may be necessary to provide the level of empathy and understanding required.

Additionally, agentic workflows may require significant investment in technology and training, and may not be feasible for smaller companies or those with limited resources. In these cases, other approaches to customer service may be more effective, such as using human customer service agents or implementing more traditional customer service processes.