As Governments across the globe struggled to meet demand for their services during lockdown, the Norwegian Labour and Welfare Administration was quick to demonstrate the way forward by using conversational AI to help its citizens to access social benefits 24/7.
Norway, which locked down early, on March 12 2020, has seen around 8,000 infections and more than 200 COVID-19 related deaths. While the Norwegian Government has already rolled out its plans to further ease COVID-19 restrictions; the spread of the virus and lockdown has exponentially increased the needs of every Norwegian municipality and its citizens.
As a result, demand on all public agencies, including the Norwegian Labour and Welfare Administration (NAV) has skyrocketed, with many citizens struggling to find the information and the help they need.
During the crisis, the Norwegian Government introduced significant measures to support jobs, help businesses and people access the help they need; strengthening social and health services and support networks. While recognising that more help will be needed in the coming weeks and months, the measures that it has put in place as the scale of the situation became clear have been rapid, targeted and effective.
One measure that stands out is how NAV is using conversational AI to help its citizens to get access to social benefits during lockdown. The agency is responsible for administering many of Norway’s key social benefit programs - including pensions, child support, unemployment benefits and employee sick leave. Despite unprecedented surges in inquiries due to coronavirus, NAV has managed to cope with a 250% increase in inquiries and unprecedented demand for its services, thanks to its virtual agent - Frida.
Speaking with NAV’s internal employee magazine, MEMU, Jørn Torbergsen, Director of NAV Contact Center, is quick to praise the work that Frida and her AI training team have accomplished. In a few short weeks, the virtual agent answered over 270,000 inquiries from citizens concerned about their situation as it related to the coronavirus.
To put this into perspective, there are approximately 850 support representatives working at NAV’s contact center, located at 15 different locations across Norway. Employees split their time between answering calls, e-mails and live chat and, at times, were receiving upwards of 15,000 calls per day. Even with the reallocation of 70 additional temporary staff from other departments within the agency, it was an insurmountable task to manage.
Torbergsen stresses that the pressure on his staff was unprecedented during this period and that Frida was directly responsible for helping to alleviate many of the challenges faced by the contact center.
During periods of extreme traffic, the virtual agent was handling incoming inquiries that corresponded to the work of 220 full-time employees - allowing NAV to give citizens the answers they were looking for quickly, and without having to drastically scale up support personnel.
The response to Frida from the public was overwhelmingly positive, with four out of five interactions successfully resolved without the need to escalate to a human service representative.
The AI Trainer team behind Frida played a critical role in the virtual agent’s success. Consisting of six non-technical NAV employees, whose knowledge and understanding of the agency’s policies and procedures made them uniquely qualified to keep Frida operating at maximum capacity.
During the crisis, the team focused on providing the best possible answers in relation to employment and health, as these were the areas in which Frida drew the most questions. As the situation proved to be fluid, the team often added new information and updated responses several times per day.
"We simply could not have done this without Frida by our side in these times."
— Jørn Torbergsen, Director of NAV Contact Center
The nature of boost.ai’s no-code conversational AI platform that Frida is built on also meant that any feedback from ongoing conversations could be quickly and easily implemented, ensuring that the virtual agent was not only always ready with the correct information, but that her tone and messaging could be empathetic, caring and easily matched to fit an individual situation.
In order to ensure informational consistency, Frida’s AI trainers worked closely in cooperation with the team responsible for the NAV website. And, as Frida isn’t subject to the same recommended work-from-home regulations as her colleagues, it was possible for the AI trainers to maintain and manage the virtual agent remotely, while coordinating over video chat, which resulted in zero downtime.