Random graph visualization

Predictive Intelligence speeds up UTM I&ITS ticket response: Interview with UTM I&ITS' Jenny Hu

Thursday, March 18, 2021 - 9:36am
Kate Martin

I&ITS is rolling out a new system to second guess you – in a good way. The ServiceNow platform, used to process tickets submitted by UTM users when things go wrong with their technology, now features Predictive Intelligence (PI). The PI function allows UTM’s I&ITS agents to fix problems with more efficiency by automating routine tasks.

UTM is the first campus in the U of T community to use Predictive Intelligence, which is being implemented under the direction of Senior ServiceNow developer Jenny Hu.

Jenny Hu, UTM I&ITS Senior ServiceNow Developer
Jenny Hu, UTM I&ITS Senior ServiceNow Developer

Hu, who was named to Acorio's 2020 list of ServiceNow Influencers, says their main focus with the PI is using Machine Learning (ML) to route incident tickets to the right group i.e., if someone notices that a light needs to be changed, the system will automatically send the ticket to the UTM facilities engineering team for response. This sorting process was previously handled by a human.

“Triaging is manual and very mundane work,” says Hu, a full stack developer with more than 15 years of experience in web application development. “This frees up staff to do other work, like writing KB (Knowledge Base) articles to help people do self-service, which also speeds up the process.”

The PI system learned to categorize by analyzing training data, made up of a select group of 20,000 tickets that have been received and closed since January 2018. “ML needs lots of data to learn how to predict which category the ticket belongs to,” says Hu, who, along with fellow programmer Selena Panchoo, started and runs UTM Women in Tech (WIT) meetups.

Although ML (a form of Artificial Intelligence) may sound like sci-fi to the average user, Hu says it is already common in our lives: Services such as UberEats use it to predict delivery times while Amazon and Netflix analyze order and viewing patterns to generate recommendations.

Since implementation five months ago, UTM’s system has already made 438 predictions on UTM Facilities incident tickets, with a success rate of 86 per cent, saving the Service Desk more than four hours a month for more meaningful tasks.

“It’s not a silver bullet, it won’t solve all problems 100 per cent,” says Hu, a U of T alum who joined the UTM IT department as a webmaster in 2003.

The key, says Hu, is that even when the PI gets it wrong, it uses that information to evolve.

"Worst case right now, the ticket goes to the wrong unit and we have to reassign it," she says, "but that means there's more data for it to learn from and is all part of training the system."

Now that PI is showing signs of success, the team is looking for other ways to use data to make predictions and generate insights that benefit the UTM community.

“We’ve only scratched the service with predictive intelligence,” says Hu.