How IBM Watson AI scales tech support for thousands

Watson saves each IBM tech support professional 45 minutes of work per day.

IBM is using Watson, the artificially intelligent data analytics processor to scale tech support to thousands of clients across 170 countries. With Watson's assistance, each tech professional in IBM's support center saves 45 minutes of work per day, said Joanna Dapkevich, vice president of IBM support tooling and Watson innovation.

SEE: How an IBM social engineer hacked two CBS reporters--and then revealed the tricks behind her phishing and spoofing attacks (free PDF) (TechRepublic)

With more than 7,000 products, IBM's tech support professionals had faced difficulty trying to assist every client with every tech issue that concerned an IBM device. Watson artificial intelligence (AI) helps organize and properly distribute this workload, Dapkevich said. 

Watson's workload 

Dapkevich outlined the following three ways Watson AI significantly improved operations at IBM's tech support center. 

1. Ticket organization

In IBM's tech support center there's 20,000 support agents. "They have a queue of tickets anywhere from 40 to 60 when they walk in. We shadowed the support agents and said, 'Other than severity, how do you approach this big list every day?' And they said, 'We read through all 40 to 60 and then we determine the order,'" Dapkevich said. 

Before Watson, tech professionals analyzed every ticket, assessing details like when was the last time the client was contacted and the tone of the client on the ticket. This process took each employee approximately 45 minutes, Dapkevich found. 

Dapkevich asked the tech professionals, "What if Watson can do all of that for you?" And sure enough, Watson did, saving them 45 minutes of reading and organization to instead work on actual client problems, Dapkevich said.

2. Ticket description

Along with organizing and analyzing each ticket, Watson also creates prompts that help clients form a description of the ticket issue, and then sends the ticket to the correct support agent, Dapkevich said. 

By using cognitive routing to help clients create their cases, tech professionals and clients are seeing a 24% to 28% reduction in time to resolution. "Before we applied these solutions, it took 10 days to solve something. Now, it can be done in seven days," Dapkevich said. 

Watson is able to read the tickets and organize them because it is trained to recognize certain descriptive words. It has been trained on hundreds of thousands of tickets to understand keywords that will help Watson direct the ticket to the right person, Dapkevich said.

3. Chatbots

One of the most sophisticated ways Watson helps tech support is via a chatbot on the support portal, according to Dapkevich.

"If you are a client, you would come to our support portal, and Watson would be embedded throughout the lifecycle [with] automation bots. Maybe you just want to reset a license key, and instead of talking to a support agent, there's an automation bot that will reset it for you," Dapkevich said. "It's done in seconds."

"Instead of saddling a support agent who's very technical and knowledgeable to do this mundane, repetitive type of work, the client gets the answer faster with a sophisticated cognitive chatbot," Dapkevich added. 

How to provide support at scale 

The key to successfully deploying a solution like this at scale involves agility and pace, according to Dapkevich.

"The biggest failures I've seen are people that have spent two or three years on something really large, and then come to find out it didn't deliver what they were hoping," Dapkevich said.

Dapkevich also recommended reminding support agents that Watson is not replacing any employees, rather, it is acting as a team member, supporting the human professional. 

For more, check out IBM Watson Media releases mobile video streaming app on TechRepublic.

Also see 

ibm-watson-munich.jpg