Creating a seamless digital support experience can make employees, service desk agents, and businesses more successful. Automating mundane, time-consuming tasks, such as triaging incidents and shifting to a proactive support model offers numerous benefits:
Increased service desk responsiveness
Reduced mean time to resolve (MTTR)
Increased incident deflection
That’s why ServiceNow and Microsoft entered a strategic partnership to help mutual customers accelerate digital transformation, create great experiences, and unlock productivity, including leveraging artificial intelligence (AI). Microsoft shares best practices and insights on how other customers can use ServiceNow® Predictive Intelligence to route issues to the right team, save agents time, and free them to focus on higher-value work.
An AI approach
Predictive Intelligence, a machine learning capability, is one component of Microsoft’s digital transformation strategy. Others include monitoring, anomaly detection, and solution prediction to create a better experience for end users and agents.
“AI makes a lot of sense when you want to automate repetitive tasks or predict future issues,” says Ryan Yee, a senior product manager at Microsoft. The company initially built an experimental AI solution that routed a limited number of incidents, allowing the team to learn and identify opportunities for innovation.
“We then took those learnings and shared them with ServiceNow, with the goal of further strengthening out-of-the-box ServiceNow capabilities so our mutual customers can start fast and move fast,” Yee explains.
By evolving from an experimental solution to Predictive Intelligence, Microsoft got the out-off-the-box flexibility and scalability it needed to drive adoption. “We wanted an AI solution that was already integrated with ServiceNow ITSM, built on our experimental work to improve our incident management process, and didn’t require extensive customization or support,” Yee adds. “That’s exactly what this approach delivered.”
Getting organizational buy-in
Although ServiceNow and Microsoft are strategic partners, it was critical for Yee and his team to get buy-in from key stakeholders at Microsoft. Rather than taking a big bang approach, he started by getting approval for a pilot program at Microsoft’s global help desk to demonstrate the potential for broader adoption in other lines of business.
Data analytics helped Yee and his team identify areas of opportunity prior to engaging the global help desk team. This enabled Microsoft to determine target metrics and define the success criteria for the initial proof-of-value exercise.
The team identified a queue that had high triage time before moving a ticket. “The existing experimental tool was only routing some of the team’s incidents,” Yee explains. “So, we proposed a pilot to route the remaining tickets using Predictive Intelligence to improve their day-to-day operations and not take anything away, which was critical.”
The global help desk team saw the potential value and helped Yee move to the next stage. “Getting your customer onboard is essential,” Yee says. “We’re a solutions team, so our job is to solve our customers’ problems. Having the support of the global help desk team gave us the credibility and impetus we needed to get broader organizational buy-in.”
Addressing data access and security is also important. Because Yee’s team needed real incidents to train Predictive Intelligence, it had to move production data into a development environment, which had security implications.
“We proactively engaged the security team to provide appropriate information and grant us approval,” Yee says. “ServiceNow helped us with this, including providing detailed security documentation. My advice is this: Put security front and center, know your data governance and security organization’s processes and policies, and address these early on.”
Start small, iterate, and scale
Microsoft’s development team created a prototype for the global help desk to try out Predictive Intelligence. This allowed Yee’s team to validate that the product could achieve the required level of routing accuracy. The prototype also gave the team proof of concept they could take to the global help desk team to quickly identify gaps rather than building a full-fledged solution that would be hard to change.
“By starting with a prototype, we had something up and running quickly, particularly since ServiceNow was there to answer our questions,” explains Chris Grumbles, a lead engineer at Microsoft. “And it meant that we could get early feedback on the out-of-the-box functionality, which proved to be critical. Our iterative approach paid real dividends as we identified key capabilities that were critical to go live.”
The service desk agents wanted Predictive Intelligence to make routing suggestions until they felt confident in the accuracy. The development team was able to customize Predictive Intelligence to write suggestions to incident work notes rather than automatically routing incidents.
The service desk team maintains control and can turn the functionality on or off. They can enable auto-routing when they feel ready. Even when auto-routing is enabled, Predictive Intelligence will only route incidents that exceed a configurable threshold, adding routing suggestions to the rest and sending them through the normal triage process.
Based on feedback from Microsoft and others, ServiceNow created an easy way for customers to configure suggestions versus auto-routing in Flow Designer, and it’s in the product roadmap to simplify further.
Planning for scale is imperative. “Starting with a prototype lets you start fast and iterate, so you deliver a better solution sooner,” Grumbles adds. Before you go live, however, it’s important to think about how you’re going to scale. For instance, what would be the impact if you needed to add new support groups or cope with new types of incidents? Planning upfront can avoid rework later.
Microsoft also found it important to make Predictive Intelligence measurable by identifying and enabling key metrics and reports. According to Yee, the process owner needs to know how the solution is performing to build confidence and trust. In addition, it’s essential for adoption—a crucial way to demonstrate benefits and bring other teams on board.
Going live and beyond
Microsoft went live with a small number of assignment groups (routing destinations) to build the confidence of the service desk team. After receiving positive feedback, the company expanded the number of assignment groups.
The rollout required minimal changes since Predictive Intelligence was automating an existing process and the service desk team was already using an experimental AI tool for automated routing. The service desk team feels in control since they can build confidence in end users by starting with suggestions and adjusting the threshold for auto-routing as needed.
“As a customer of this solution, we want our agents to focus on issue resolution for the user and not having to waste time on incident classification, routing, and basic triage activities,” notes David Finney, a principal IT service operations manager at Microsoft. “We want issues getting to the right agent the first time all the time. Automating these tasks is critical to our success in reducing resolution time for the user. Predictive Intelligence provides us with the capability to implement this."
Three key considerations help with the process and cultural changes that are inevitable when adopting AI:
Set realistic expectations around AI. Although AI, including machine learning, is good at solving specific problems, it’s not a magic bullet. It needs to learn. You need to manage Predictive Intelligence just like any system, and you need to recognize it will “learn” and improve over time.
Data selection and quality are critical. Predictive Intelligence excels at handling inputs such as employee-reported incidents. But if you train it with bad data, it will negatively affect the model. In addition, you should understand your as-is process around incorrect assignments (human agents can make mistakes) so you can establish a baseline and compare. However, after you go live, it’s important to monitor routing accuracy and look for corresponding data issues when accuracy degrades, to ensure you maintain high-quality data.
Collaborate with the process owner of the target team to align with their business imperatives. For example, Yee’s team held off going live until the global help desk had completed its financial end of year.
Microsoft is already seeing significant interest in Predictive Intelligence from other lines of business. Based on the company’s successful experience, Microsoft and ServiceNow identified six key recommendations for other organizations looking to implement Predictive Intelligence or Predictive Intelligence Workbench (which presents AI solutions as use cases):
1. Identify which data you want to use and review or filter accordingly, including for quality (e.g., are all fields filled?). Engage your security team early to address data governance and security requirements.
2. Communicate the benefits and work with process owners to get broad organizational buy-in and approval. Set realistic expectations with your stakeholders about what AI can—and can’t—do.
3. Start with a pilot to validate the solution and get iterative end-user feedback. For example, start small with a single team (even if you have multiple support desks) and begin with routing suggestions before transitioning to automatic incident routing. That way, you can collect end-user feedback, learn, and understand best practices to scale.
4. After you go live, monitor your solution for accuracy and look for corresponding data issues. Machine learning requires ongoing ownership to ensure the solution is performing.
5. Leverage your initial implementation to create champions and back up your success with hard data. A key part is making sure your solution includes metrics and reports to build confidence and prove value.
6. Focus on the overall experience, not just the technology and functionality. Keep in mind that success is about empowering people, not just deploying AI technology.
Find out more in Modernizing the support experience with ServiceNow and Microsoft.
Want to participate in user experience research to help make ServiceNow products better? Visit the ServiceNow Product Lab and become a product lab insider.
© 2022 ServiceNow, Inc. All rights reserved. ServiceNow, the ServiceNow logo, Now, and other ServiceNow marks are trademarks and/or registered trademarks of ServiceNow, Inc. in the United States and/or other countries. Other company names, product names, and logos may be trademarks of the respective companies with which they are associated.