When we talk about training a bot, it first seems like a mechanical and impersonal task. But it’s nothing like that. With all the tools it has, it’s possible to personalize the service and thus achieve an efficiency rate of over 90%
Mônica Batista Vargas de Castro - CIEE Operations Manager
Is it possible in two weeks to migrate 600 agents to remote work, include an automated service solution and maintain the already recognized quality of the customer experience? This was the case of the Centro de Integración Empresa-Escola - CIEE, a civil association governed by private law, non-profit and non-economic purposes, whose objective is to help young people and students to access the world of work through internship programs and learning throughout Brazil.
In March 2020, the beginning of the COVID-19 crisis in the country, it created an uncertain situation for CIEE and, considering the health of its team and the maintenance of its work, the conversational AI
implemented in record time was the answer to continue offering high-level service. Between March 25th and May 31st, more than 42,600 bot sessions were recorded, with an average attention time of 1 minute and 13 seconds, and a response efficiency of around 90%.
And to address special issues that require human participation, live chat with AI for agents was also included in CIEE's routine, receiving 78% positive feedback from users in its first month of operation. Launched one week after the chatbot, between 4/2 and 5/31, over 4500 sessions were present, with an average attention span of 1 minute and 32 seconds.
A change in strategy
Conversational AI arrived at CIEE in the hectic context of the pandemic and, as in most companies, the health of employees was prioritized. And, given that the main audience for customer service channels are young people and students (that is, millennials and centennials), the migration of services to digital channels was very well received.
In the words of Mônica Batista Vargas de Castro, CIEE's Operations manager: "I think digital channels are here to stay. They were already a trend that was going to settle naturally, but the pandemic led many companies to anticipate and implement it in their action plan. And I think it’s a very efficient way to give customers new experiences"
The entire process of automation of digital channels and migration to remote work took approximately 15 days, including the qualification of the team for content management, training and implementation of the bot.
And a great challenge was to automate the service without neglecting the human, personal and empathic factor that has always been part of the experience offered by CIEE.
Therefore, there was a continuous monitoring of customer service throughout the process, looking at each step of the interaction to improve the bot's content, enabling a more accurate connection with users.
Metrics and reports: The key to improving the future
Accessing the metrics is the key to assess the bot's performance and discover the most important insights from its interactions, something that has had a very favorable result for CIEE. When it comes to artificial intelligence and customer service, good results are not achieved by magic, but as a consequence of a team dedicated to understanding what to improve and how.
Before the pandemic, students could only contact the organization on predetermined weekdays and times. But the bot showed the huge demand of its public to connect in its own time: A good part of the interactions since the beginning of the bot operation happened on Saturdays and Sundays.
This meant that CIEE could provide support outside business hours, immediately and 24/7, adapting to the way students communicate - and not the other way around. Thanks to the conversational bot, students don't have to wait until Monday to get an answer.
Another interesting point was the work to shorten the service times, since we live in a context where the speed of response is more valued than ever. And the average service time of CIEEs’ bot was about 1 minute and 13 seconds in the first 2 months.
In addition, only 10% of the consultations were transferred to the service agents, demonstrating that 90% of the users' questions could be automated and answered immediately by the bot, also showing the high degree of adaptation of their audience.
We're really satisfied with these numbers.This information is very significant because it shows we made the right decision and that we're on the right track, Monica explains.