After-contact work is one of the most important factors to consider when discussing contact center efficiency. According to the International Finance Corporation, agents take an average of 6 minutes to complete the case once a call is finished. During this time, common tasks include documenting the call, categorizing the call outcome, escalating issues, or updating customer information.
These tasks are crucial for ensuring a high level of customer service, compliance with regulations, and overall success of the contact center, but they might hurt return on investment (ROI) by taking agents off the line.
Large Language Model (LLM) capabilities can dramatically reduce time spent performing any documentation and categorization tasks while keeping a high operational standard. Copilot transcribes the call in real time, and by the time the call is finished, it can provide:
- an automatic summary of the conversation, that the agent can copy to other systems such as the CRM tool.
- a list of the next steps that were agreed upon with the customer, so that agents can quickly take action.
- a recommendation of which disposition label is best suited considering the content of the conversation.
How it works
These features are provided with your Copilot subscription.
Please follow the steps below to make use of them:
1. Once the agent enters the call wrap-up stage, the Copilot app will automatically update to show a different screen, which includes the “Disposition and Summary” section [1], where the agent will see the Generate with AI [2] button.
Note: Please note that to see these options on Copilot, you need to enable After Call Work on Talkdesk.
2. Then, on the Copilot application, agents will be able to see:
- The recommended disposition (this is selected from the existing disposition labels in your account) [3].
- The summary of the interaction [4]. You can leverage automations to share the summaries with your Talkdesk Connections. Learn about automations on the Creating an Automation with Talkdesk Connections article.
- The list of agents' next steps [5].
3. On the same screen, agents will also be able to use the tools available to:
- Copy to clipboard [6].
- Add to notes [7].
With these options, the content of the summary and next steps is copied to the “Notes” widget in the “Conversations” app. In addition, the recommended disposition in the “Disposition” dropdown in the “Conversations” app will be pre-selected.
Note: If you don’t use Talkdesk “Dispositions”, you can still take advantage of “Summarization and Next Steps”. If no disposition is created on your account, Copilot will not generate a disposition recommendation, but it will still generate the summary along with “Agent next steps”.
If you would like to automatically log the call summary in a third-party application, we recommend you add the content to the “Notes” widget and then use Talkdesk “Connections” to automatically sync “Notes” with your system.
On-Demand Summary
The On-Demand Summary feature empowers agents by providing instant summaries of ongoing conversations. This allows agents to quickly grasp key discussion points and the conversation's context without reviewing the full transcript, leading to more efficient and focused interactions. To enable this feature, administrators can navigate to AI Launchpad > Global properties > Summary. Once enabled, agents will have the option to generate a summary during a live conversation.
A minimum of 1 new utterance in the conversation is required before a summary can be generated, and 20 seconds cooldown period between requests for a new summary on the same conversation. Additionally, administrators can configure an event within Integrations that triggers when an on-demand summary is ready.
Fine Tuning
Fine Tuning enables you to customize the default summarization prompt instruction used by Copilot. To start using this feature navigate to the Use Cases section on AI Trainer.
With this, users can adjust how the content is summarized to better align with their needs. The main areas that can be customized are:
- Formatting: Control how the summary is presented. Whether you prefer bullet points, numbered lists, or structured paragraphs, you can modify the format to match your preferences or organizational standards.
- Length: Adjust the level of detail in the summary. If you need brief, concise summaries or more comprehensive overviews, the Fine-Tuning feature allows you to set the summary length that best suits the task at hand.
Fine Tuning Simulator
The Fine Tuning feature includes a Simulation of Summaries to help you understand how customizations will impact your summarization. This simulation allows you to test the changes you’ve made and immediately see their effect on a previous interaction’s summary.
Here’s how it works:
- Select a previous interaction and its generated summary [1].
- Apply your customizations [2].
- Compare the original summary with the customized version [3].
To ensure clear and concise instruction, always specify exactly what information is required and the structure you expect. For instance, if summarizing a customer interaction, ask the AI to include the customer’s name, phone number, and address, as well as the purpose of the call and its resolution. When specific formatting is needed, provide a simple example. Learn more here.