Rapid deployment of 311 virtual agents helped address increasing call volumes
The City of Sacramento implemented virtual agents to handle incoming calls for Sacramento 311. The single point resource connects with the City of Sacramento for government service requests for building permits, stray animals, utility services, and much more. The resource is a 24×7 operation handling approximately 1,500 non-emergency service requests from Sacramento residents, visitors and businesses per day. Given this is a high volume call center, citizens experienced longer wait times to get the right level of support.
“Our goal with this project was to try AI based solutions to solve real world problems and provide better customer service to our constituents. We see this project as the beginning of using and integrating AI & ML technologies to transform the business process at the city.”
Instant engagement using virtual agents with context aware support
Callers trying to reach the City of Sacramento 311 sometimes experienced longer wait times due to limited staff which led to frustration, and a less than ideal customer experience. Calls received at 311 continue to increase year after year and the recent pandemic added more stress on the overall system.
For the past couple of years, the city was laying the groundwork for the complete renovation of the city’s 311. Part of this initiative included implementing a new CRM system, updating the phone system, workforce management, and introduction of Artificial Intelligence (AI) solution to augment the workload of 311 agents.
End to End seamless integration: Google Cloud Dialogflow integrated with Cisco IVR, Salesforce and GIS systems
Google Cloud Dialogflow was implemented as the virtual agent to interact with citizens. Depending on the nature of the requests the virtual agents transferred the call to live agents using Cisco IVR. All the performance metrics on call volume, handling time, effectiveness of the virtual agent were captured in BigQuery and visualized in Data Studio. This helped to understand the call request intents and provided inputs to continuously train the virtual agent. Additionally, the solution was integrated with Salesforce and GIS systems for contact verification, contact creation, address validation, and address suggestions to enrich caller case data & record creation with the call details from the virtual agent.
AI for scale and human in the loop to ensure high quality service
Within the first month of going live, the virtual agent performs better than a human agent by handling more calls per day. This project phase focused on developing integration with core systems supporting the 311 operation and handling simple use cases.
Before the implementation of the new solution, citizens would go through long wait times which caused frustrations and delays. By verifying the callers’ service address and intent, the new system can reduce the overall call time and connect to the most qualified agent available at that moment.
With the GIS integration, the new system can confirm users’ intended service address through the voice conversation and make sure it is within city limits. The system can redirect calls not serviced by the city to other local agencies like SMUD, PGE, and County, resulting in faster response, better service and reduction in overall wait time.
Additionally, if the virtual agent is not trained on a specific skill or is unable to understand the user, the system is able to transfer the call to a human agent along with conversation data to complete the service request. We strongly believe a well trained virtual agent helps to provide instant support and combining with human in the loop enabled to have high quality customer services. The City of Sacramento 311 use case is a great example of utilizing AI to serve the citizens.