The IT helpdesk of a major US technology company receives thousands of technical questions each day. Around 70% are repetitive. Responding can take days, stopping employees from achieving their tasks faster.
Most of the questions include a lot of technical and company-specific terms which are hard to pick up for SaaS-based NLU systems. Furthermore, these interactions include sensitive data which cannot leave the organization due to data regulations.
The company deployed a conversational AI system on-prem using the Rasa Platform. Having thousands of historic chat logs, the company trained their own, domain-specific language model which increased accuracy significantly.
The conversational AI reduced the average response time by 80% and increased job satisfaction of IT helpdesk staff and employees. The solution will now be introduced to all departments within the company, handling thousands of requests each week.