eWEEK content and product recommendations are editorially independent. We may make money when you click on links to our partners. Learn More.
Due to a number of different circumstances that have come together in the last half-dozen years to form an important convergence, artificial intelligence and machine learning are becoming more portable for use at the edge, in addition to their usual homes in the data center or in the cloud.
It’s all here now: high-speed bandwidth, 5G connectivity, super high-quality code and code libraries, unprecedentedly powerful processors that use less power than previous models, unlimited storage capacities, ingeniously designed mobile and stationary connected devices, a zillion types of cloud services–we could go on. What is next?
We’re already seeing it. the introduction of more functionality through artificial intelligence. We’re seeing more AI in more apps in more places than we’ve ever seen before: wearables, cars, productivity apps, military, health care, home entertainment–the list is lengthy.
This question-and-answer article is conducted with topic expert Vaibhav Nivargi, CTO and co-founder of Mountain View, Calif.-based Moveworks, which works on the front lines each day helping companies on their inclusion of AI for use in a wide range of IT use cases.
Q: You hear about AI impacting a number of industries, but IT support seems like an underserved area. Can you explain that?
Nivargi: IT support is—for the most part—still managed by a team of people, which means the process tends to be painfully slow. Across industries, we’re seeing the average IT support ticket take three business days to resolve, bringing productivity to a standstill at a time when we’re fully reliant on our technology to get anything done. And because IT teams are overwhelmed with fixing routine tech issues, like resetting passwords and editing email groups, they lack the time to focus on critical digital transformation projects.
Beyond its effects on productivity and digital transformation, the manual approach to IT support also directly impacts the bottom line. Consider this: an average employee submits one IT ticket per month. And on average, each ticket costs about $25 to resolve using conventional means: service desk agents, workflow tools, service centers, and so on. That’s around $3 million spent just on support costs per year for a company with 10,000 employees.
What we’ve done at Moveworks enabled IT teams to resolve these tickets automatically with AI, at a fraction of the cost. To your point, artificial intelligence and machine learning have helped to accelerate and scale a huge number of processes across all lines of business. Extending that logic to supporting employees at work is now producing similarly powerful results.
Q: IT has been pushing to modernize for quite some time. What’s been holding it back?
Nivargi: In our experience, CIOs at many large companies—who really should be advocates for modernizing IT—are concerned about adding any new AI solutions. They have two main objections. For one, they worry that adding more tools will increase complexity for end users, and for another, they fear that maintaining AI will require so much work for their team that the end result will be net negative. Of course, these are valid considerations. Most of us are already using more tools and applications than we’d like, and many AI solutions are in fact “AI toolkits” that IT teams must build and manage themselves.
It’s interesting to get this pushback against IT modernization, since the whole point is to reduce the complexity for end users and the burden on the IT team. That’s why our approach is to completely handle our chatbot’s language processing models and back-end integrations, so that the behind-the-scenes tech is invisible to employees and agents. Ultimately, modernizing IT means simplifying IT: allowing employees to get the support they need automatically by talking to a bot in normal conversation, without involving the IT team.
Q: This was an interesting year for IT given they had to keep a remote workforce productive. What was that like for you and your customers?
Nivargi: We never anticipated how extensively our platform would be used in a work-from-anywhere scenario. Virtually overnight, the IT service desk became mission-critical to every department inside organizations across all industries. Getting tech support is now more urgent than ever, with employees needing immediate help no matter where they are and when they ask. It’s a problem tailor-made for AI.
On the service desk side, the initial transition to working from home flooded agents with high-volume requests, while new challenges—such as keeping remote employees up to speed and answering questions about company policy—seem like they’re here to stay. At one of our customers, Unity Technologies, the first month of work-from-home was hectic, to say the least: requests for Zoom licenses increased 6X, questions about policy increased 5X, and the overall number of IT issues nearly doubled.
Unity’s workforce responded by turning to their Moveworks chatbot, which they’ve nicknamed Ninja Unicorn, to get help. There were three times as many interactions with Ninja Unicorn during that shift to remote work, and the result was Unity’s IT team successfully meeting the increased demand for support.
The future of work is very much uncertain, but there’s no doubt that AI will play a central role.
Q: Moveworks intersects with some interesting technology trends. I am curious about the AI, given IT departments tend to have smaller data sets: Doesn’t AI need large data sets to produce results?
Nivargi: Great question. It’s true that effectively training a machine learning (ML) model requires “big data” — that is, millions of data points that ultimately allow the model to make confident predictions. The problem is that most companies simply don’t have that many examples to use. Google, for example, has 30 trillion web pages to refine its algorithms, whereas a smaller company attempting to automate an internal process might only have 30 relevant data points to train a model.
But if we can overcome its obvious shortcomings for ML, “small data” in the enterprise is actually quite useful. Experts point out that small data has more potential for individual relevance than its big data counterpart, for example, since small data can capture patterns and trends from an extremely specific context.
Thanks to bigger data sets, budgets and ML teams, big tech companies still have an advantage. But fortunately, nearly any organization can put AI to work with minimal effort—and without hiring a dedicated team of experts to maintain it. Using techniques such as Collective Learning, third-party AI vendors can support medium-sized organizations and help them benefit from powerful ML. This is particularly true for use cases in which uniform patterns emerge in data from across multiple companies, such as IT support.
In short, ML efforts don’t have to be hampered by the problem of “small data.”
Q: Do you have any parting advice for IT execs and CIOs as they think about accelerating IT support in 2021?
Nivargi: Yes, absolutely. Service desks need to stop the vicious cycle of IT support, in which agents are so busy resolving individual IT issues that they lack the time to fix underlying problems and inefficiencies, causing more issues for end users.
Of course, that’s easier said than done. Here are some steps IT teams can take to get started:
- Collect and synthesize data from across all IT tickets to spot trends in user behavior. Consistently tracking all tickets—either manually or with an automated system—yields the visibility needed to prioritize the most pressing use cases for AI.
- Identify the most common and time-consuming support requests that are potential candidates for automation; for instance, unlocking user accounts, provisioning software, editing email groups, and ordering devices.
- Implement an AI chatbot that can resolve these common requests, without service desk intervention. Some chatbots, including that of Moveworks, also provide visibility over user behavior trends to accelerate steps 1 and 2.
One last point: Several of our customers feared their environments weren’t mature enough for AL and ML when we first partnered with them. Ask them now, and they’ll tell you the secret was simply to begin the process. You often don’t know what’s broken until you see the solution, and particularly with machine learning, that solution is an appreciating asset that becomes more impactful over time. Getting time on your side means starting the clock.
eWEEK is testing “Industry Perspective” as a beta-type feature. Let us know your opinion of it at cpreimesberger@eweek.com.
