BACKGROUND IMAGE: iSTOCK/GETTY IMAGES
At its Knowledge17 conference earlier this month, ServiceNow Inc. debuted some cloud-based applications, including several that pit the ITSM technology against some formidable competitors in new markets, such as customer service management and human resources.
Farrell Hough, vice president and general manager of IT systems management, with oversight of product operations at ServiceNow, based in Santa Clara, Calif., spoke with senior executive editor Ed Scannell about the ServiceNow roadmap, the role of artificial intelligence (AI) and machine learning in IT operations, and their addition to ServiceNow's flagship Intelligent Automation Engine.
These latest web-based applications are an aggressive push into several new markets for the company. What is the overall strategy here?
Farrell Hough: In the IT market, there has been a clear indication people are interested in adopting AI because the benefits are tangible. There are a lot of point solutions out there now, but we intend to fully drive automated intelligence directly into our platform [the Intelligent Automation Engine].
Our preference is to start out with some practical steps, like introducing auto classification and categorization routing. This will eliminate administrative triage that has plagued our IT service desk admins today, along with their end users. When the categorization is correct, it ensures incidents sent my way are truly meant for me and my expertise. The time and level of investment that goes into administrative triage on the front end is enormous.
What AI and machine learning capabilities are in the new Intelligent Automation Engine?
Hough: The operations intelligence component, which provides the anomaly detection, is a machine learning algorithm. The auto classification component comes directly from our replatformed version of the DxContinuum acquisition we made, which is an AI-based algorithm. The predictive analytics component of our performance analytics package has algorithms in there to get you more into forecasting that is not just linear. The benchmark component is not machine-learning-based, but we want to take that there so it can make recommendations based on performance and make suggestions for what else you can add in.
Besides the Intelligent Automation Engine, what AI and machine learning technologies have you built into the new applications?
Hough: First, each component of the Intelligent Automation Engine can be applied to any of our applications. With the case management for HR or the customer service management applications, they face the same struggles with administrative overhead as other applications do. But the new applications can grab a number of AI and machine learning capabilities from the platform [Intelligent Automation Engine]. The same is true with the benchmarks application. The next technology we will go after will be chat bots and virtual agents, which is not in our platform today.
Dave Wright [ServiceNow's chief strategy officer] suggested you don't want to put too much in-the-box AI capabilities into a platform or application -- to leave some room for users themselves to innovate. Where do you draw the line?
Hough: We started out with phone scripts and IVR [interactive voice response] trees that we pushed out to users to ease workflow. But what we were really trying to do is get humans to function more like machines. That continued as we built out our extended knowledge engineering, turned them into knowledge bases and made them searchable on the web. Now, we are building a self-service portal that allows them to search for service catalog items, so they can find things on their own.
We optimized for the service desk efficiencies, but unfortunately, I don't think we optimized for the user experience. But we are now at this inflection point where we can do both. From here on, there won't be any tolerance for not taking into consideration what it means for the user experience. You will not be given a second chance if you don't do it right in deploying machine learning capabilities. You have to take it slow when you are deploying [machine learning technology] and allow for users to opt out of whatever you are putting in front of them.
How can users innovate with AI-related technologies, given that many IT shops don't have enough trained personnel who can effectively work with AI?
Hough: There is a skills gap that will require investing in STEM [Science, Technology, Engineering and Math], and not just in the tech industry, but on a larger scale in our country and globally. We have time -- this is not a next year kind of initiative. But working with AI and machine learning will require new skills.
CIOs and CTOs understand the amount of automation that will be required in the future, if only because of the sheer amount of data they will have to deal with from IoT [internet of things] and mobile devices. But the good news is, IT workers belabored by the level of administrative triage -- i.e., manual tasks -- can offload all this redundant, mind-numbing work and have the chance to reskill on the job. It is a new form of knowledge engineering.
When IT workers hear about automation, they break out in a cold sweat, afraid their jobs will be eliminated. What are the realistic prospects for retraining all these people that do mundane IT work?
Hough: IT workers are knowledge engineers and have understood knowledge engineering for years, whether it was putting in a phone tree or an IVR system. So, now, they would be designing conversations for machines. I would say inside of five years, machines will be able to automatically generate these conversations. But we will need resources that can understand the context to be able to curate, manage and evolve what those conversations will look like. There are job titles out there on LinkedIn that are starting to pop up, such as conversation designer, specifically shaped to do this.
The new applications bring you into new markets with large competitors like Oracle and SAP, as well as smaller but nimble web-based application companies. Does this give you concern?
Hough: I think the power we have in the platform creates the differentiation for our applications. There are lots of point solutions out there for security and HR and CRM [customer relationship management] apps, for instance, but there [aren't many] of them that are on a platform that connects to the parts of the business a company heavily relies on.
How much can you lean on a partnership like the one with IBM to help you compete against large competitors like Oracle?
Editor's note: ServiceNow and IBM recently expanded their relationship to incorporate IBM's cognitive technology into ServiceNow's IT services automation platform.
Hough: IBM has been a managed service provider for us for years. They rely on our technology to be able to go across their customer base and deliver better service management. They also have complementary technologies they can offer on the cognitive side. There are some users that already have a strong relationship with IBM and are poised to use Bluemix services and leverage IBM Watson services. And we are perfectly positioned to partner and have integrations with IBM Watson services, which can help our own customers on their automation journey. And some of that will come from us, and some will come from IBM.
Some of these future AI and machine learning products, how will they be able to take full advantage of IBM's Watson?
Hough: We have partnered with several users to design ServiceNow and IBM Watson integrations. It has primarily involved using their virtual agent service, so ServiceNow can serve like a front end to IBM's software, making it easier for IT to offload some work. It has proved valuable, in some use cases, for things like checking the status on a range of different incidents, including requests that flood into a company's service desk.
Ed Scannell is a senior executive editor with TechTarget. Contact him at firstname.lastname@example.org.
If Moneyball could see machine learning and AI now
Machine learning tools translate eBay's product listings
AI improves business productivity despite its challenges