Figure 1: How AI-powered dashboards can bolster technical support for national health platforms.
At Everwell, we’ve seen firsthand how challenging it can be to support large-scale national health platforms. With thousands of questions and requests coming in from the field, technical support teams face a daunting load. To help ease this burden, we developed an AI-powered dashboard that streamlines the process. The dashboard uses large language models (LLMs) to summarize user reports, prioritize them, and link them directly to the underlying workstreams of developers.The platform is general and could be applied in a variety of settings. In India, we’ve used it to help provide technical support for Ni-kshay, India’s national tuberculosis platform, which manages over 60 million case records.
The Challenges of Technical Support
Many large-scale software services, including those supporting public health initiatives, use multi-tiered support systems. If a patient, staff member, or anyone else encounters an issue, they can submit a support request. The first tier (L1) typically consists of a call center, where many questions can be answered immediately. But if the issue requires more technical expertise, it gets escalated to a second level help desk (L2) through a written request, often called a ticket. If even more technical investigation or software modifications are needed, a ticket is created for the third level (L3), which is handled by software developers.
We wanted to dig deeper into the specific challenges faced by L2 and L3 teams, especially when dealing with large volumes of tickets. Here's what we found:
Understanding user requests: L2 tickets are often submitted with unclear or incomplete descriptions, making it difficult for support teams to quickly grasp the core issue.
Linking user issues to software problems: L2 tickets represent what the user is experiencing, while L3 tickets capture the underlying software bugs or feature requests that need to be addressed. Manually mapping these tickets to each other (across systems like Jira for L2 and Gitlab for L3) is time-consuming and can easily become overwhelming.
Prioritizing issues: Prioritizing software problems based on how many users are affected can take hours without automation, as it requires constantly reviewing and linking multiple tickets.
How An AI-Powered Dashboard Addresses These Problems
To tackle these challenges, we built an AI-powered dashboard with several key features designed to save time and boost efficiency:
AI-suggested titles and descriptions: LLMs help summarize user-submitted tickets by suggesting clearer, more detailed titles and descriptions, making it easier for support teams to understand the issue at hand.
Ticket mapping: The dashboard enables linking between user-reported issues (L2 tickets) and software issues (L3 tickets). It uses APIs to interface with systems like Jira and Gitlab, and it counts and displays how many tickets are linked, helping developers prioritize the most pressing issues. The dashboard also suggests potential software tickets that could be related to a user’s problem, using a combination of vector embeddings and LLMs.
Visual tools for prioritization: To help teams see the bigger picture, the dashboard includes visualizations that track the trends of linked tickets over time. For instance, a simple histogram shows how many users have reported issues tied to a specific software milestone, helping teams decide where to focus their efforts.
Real-World Impact
The dashboard is already in use, including as part of the support process for Ni-kshay, India’s national platform for tuberculosis. Hundreds of tickets across L2 and L3 are managed in the dashboard, and early feedback suggests that the ability to maintain and visualize the links between user issues and software tickets is transformative. Support teams have reported significant time savings, as the manual task of tracking and linking tickets is now streamlined.
We’re excited about the future of this tool and are committed to continually refining it based on user feedback and evolving needs. If our approach resonates with challenges your organization is facing in managing large technical support processes, we’d be delighted to connect and explore potential synergies.
Acknowledgments
Key contributors to this project include Manoj Muralidhar Bhagawath, Bharat Agarwal, Venugopalan Iyengar, Tulsi Bhikadia, Ganesh Shirpure, Soumabha Ray Chaudhuri, Anurag Sharma and Bill Thies.
Comentários