AI helps solve problems of 22 thousand Adobe employees

Adobe has implemented AI that can independently answer employee questions and direct their support requests in the right order of urgency. This reduced the waiting time from 10 hours to 20 minutes.

Adobe has introduced an AI model that helps freelancers get quick help with frequently asked questions. To do this, they first created a Slack channel that connected all employees of the company, but initially, the global IT support service answered questions there. But if the problem did not occur during business hours, employees had to wait for an answer for several hours.

By focusing on the most common and important questions, the company decided to optimize support for FAQs and issues. So they created an AI-powered by machine learning, it has access to all messages in the Slack channel, and it works with a built-in natural language processing (NLP) tool.

The AI ​​answers questions in plain text or posts a link to a knowledge base article or FAQ. The company initially decided to focus on the eight most frequently discussed topics, but they continue to add problems and topics – by the end of the year, they want 50% of problems to be solved with AI alone.

Adobe said that the automated system answered more than 3 thousand questions, which several times reduced the waiting time for solving urgent problems. With an NLP-based deep learning and routing engine, 38% of e-mails are now automatically routed to the correct department. The AI ​​routing bot uses a neural network classification technique to sort emails.

This AI improvement has reduced the average time it takes to resolve problems from 10 hours to less than 20 minutes. Continuous learning has helped the company achieve 97% accuracy – almost on par with living experts. As a result, the volume of requests for internal support decreased by 35%.