Empowering Call Center Agents: A Feedback-Driven Semantic Search System
Struggling with usual in their industry high turnover, our client asked us to help retain institutional knowledge and improve onboarding. We created tooling which boosted response time and quality, made knowledge more accessible and quickened new employees training. All without overloading the workforce.
The Challenge
Mid-sized call center serving multiple clients struggled with employee retention and training. While trying to address those challenges, they identified causes to be beyond their control. To handle the situation as is, they decided to improve their employee training systems.
Scope and Direction
We learned about their processes and determined points where new employees underperformed. We then identified what information was missing, how was it currently accessed and what could improve the speed and quality of its acquisition. Finally, we created a feedback system to maintain accuracy of the information and improve its quality over time.
Stage One: Providing New Employees with just-in-time Knowledge
Existing solution for employees trying to respond to customers presenting them with unknown problems was to consult Q&A database provided by their clients or fallback to a more experienced coworker. The tool provided multiple search mechanisms and was well structured, but didn’t meet ergonomic needs of employees. We replaced the website interfacing database with a browser extension which defaulted to semantic search over the question part of the database. Now employees could select customer questions and instantly query the database within five keystrokes. This sped up the process, limited context switching and provided us important buy-in from employees, which was to be critical in implementing the feedback system.
Stage Two: Integrated Feedback Mechanism
Once we acquired control over the interface, we were able to assure improving the quality of answers over time. The previous system allowed employees to issue corrections to existing answers, but it was cumbersome and rarely utilized in practice. We provided employees with a three-tiered feedback mechanism: first, we were able to check which of the provided answers they picked. Secondly, we created a simple rating modal for chosen answer, occurrence frequency of which our client could control. Finally, in case provided answers were wrong or insufficient, employees could add it to the database with the same ease with which they were able to search it. Popularity and rating of answers created by each employee was tied to their account and allowed the client to use in their motivational programs
The Impact
Stage one had minimal positive influence on speed and was neutral for its quality measures, but improved employee experience by an average of one point on seven point Likert scale. We ran a follow-up survey three months after full implementation of stage two. Internal response quality measures increased by over 10% across the board, with new employees reaching old standards in half the time. Additionally our client reported that new motivation scheme our system enabled allowed them to better retain their top employees.