Leveraging AI to Enhance Zendesk Analytics in Workforce Management

Last week, I delved into enhancing the analytical and data engineering capabilities of workforce management teams. Today, I'd like to share a practical example of how we addressed a specific challenge.

At DraftKings, we rely heavily on Zendesk as our customer ticketing platform. It enables our frontline associates to efficiently manage tickets by navigating through assigned views. Over time, our team has become proficient at using Zendesk, optimizing the platform for routing and balancing the workload across different views.

However, we encountered a hurdle: Zendesk lacks a straightforward method to review historical data on the queue size of each view (i.e queue). This limitation poses several challenges

1. For comprehensive capacity planning in a ticketing environment, it's crucial not only to know the volume of new tickets within a certain period but also to understand the backlog of tickets at the beginning of that period. Combining these two metrics provides a more accurate representation of the workload facing a team. Without the ability to examine past queue sizes, we lose a significant portion of the data needed for accurate forecasting.

2. Operators rely on standard procedures to manage ticket volumes across views and queues. The absence of historical data on queue sizes makes it difficult to assess whether these procedures are being followed correctly, hindering our ability to provide feedback on both positive and negative behaviors.

To overcome these challenges, I developed a script that connects to Zendesk at regular intervals, retrieves the number of tickets in each view, and records this data in a simple database table. This solution enables us to access historical ticket counts, addressing our need for retrospective analysis.

Creating this script was straightforward. I leveraged ChatGPT to generate a Python template that interfaces with the Zendesk API to fetch the counts. See the code example below:

The integration of generative AI and programming has significantly expedited processes that would traditionally take weeks or months, reducing them to a matter of minutes. This efficiency underscores my belief in the unprecedented ease of upskilling our WFM team, thanks to the widespread availability and user-friendliness of modern software and tools.