ServiceNows AI chief on solving the last mile problem for enterprise AI

ServiceNow’s AI chief on solving the ‘last mile problem’ for enterprise AI Derek du Preez Wed, 12/02/2020 - 05:59

Summary:
Vijay Narayanan, ServiceNow’s Chief AI Officer, explains how the workflow vendor is making AI scalable in the enterprise.

Earlier this week workflow cloud vendor ServiceNow announced that it had signed an agreement to acquire Element AI, a Canadian company that aims to apply modern AI to text and language, chat, images, search, question response, and summarization. This will mark the fourth AI acquisition ServiceNow has made in the last 12 months (following Loom Systems, Passge AI and Sweagle) and the seventh AI acquisition over the past three years or so.  Not only this, but ServiceNow also hired a new Chief AI Officer - Vijay Narayan - in March. Narayan blah blah blah  Given the acquisition activity and that Narayan is now a few months into his role, it seems like an opportune time to ask: what is ServiceNow’s strategy when it comes to implementing AI across the Now platform? We got the chance to sit down (virtually) with Narayan, who gave a rather frank assessment of the state of AI in the enterprise in recent years.  He said that whilst the capabilities are there in theory, and have been improving drastically in recent years, the enterprise community by and large hasn’t been able to translate this into effective outcomes for users and customers. Narayan explained:  Machine learning has had tremendous advances on the research side for the last 10 years or so, it’s had incredible advances. But on the enterprise side, when it comes down to applying and leveraging machine learning for driving real business value, I feel that it has not been as effective. Machine learning uses data and trains models to come up with a set of recommendations and tells you what to do next - it gives you little pointers or a set of recommendations - but most enterprises struggle to translate those decisions and recommendations into real actions.  This is something that I call the last mile problem for enterprise AI, because the machine learning platforms are typically outside of where the work actually happens. Connecting those two pieces, integrating them, setting up these fully automated systems that are training models, driving the actions that deliver business value, becomes very hard for most enterprises. 

It is the potential to solve this ‘last mile problem' at ServiceNow is why Narayan joined the vendor earlier this year. It is Narayan’s view that by implementing AI capability where the work is actually happening - the workflow - companies will be able to take huge leaps forward with efficiency, experience and agility. He added:  So when I learned about ServiceNow, that was when it became very, very exciting to me. Because this is a place where work happens in the enterprise. If you are able to bring in high quality AI capabilities into the workflow platform, you sort of close that last mile. It gives you great opportunities to go from predictions and recommendations, to actually taking actions and closing the loop. If you are able to deliver that in a single platform, with one data model, one architecture, I feel that could be very, very powerful in realising the value of enterprise AI at scale. 

Finding the use cases Narayan said that with this in mind - specifically thinking about making AI scalable in the enterprise - ServiceNow is very focused on building purpose built AI. He is aware that the term ‘AI' has been overhyped and is often bought in for the sake of it, rather than to solve problems that would make a difference. As such, Narayan and his team are thinking about the capabilities that ServiceNow has acquired over the years and how they can be applied to tangible use cases. He said:  We need to make AI very easy to consume, without having to need a lot of data scientists or machine learning engineers. This is the beauty of what I’m excited about - can we bring those AI capabilities for higher level, complete use cases, and bring them natively into the workflow platform? 

So what does that mean across the Now platform? Some examples of how AI is being used include using natural language processing to understand case requests, having them automatically assigned to the right group, and then allocating them to the right agent for resolution. ServiceNow is also using virtual agents so that companies can ask questions such as ‘How many incidents in the last 30 days have missed their SLA for solution?', and have the data returned automatically. Another use case is using machine learning to analyse machine data, correlating to events from different machine services (such as applications, infrastructure) to understand the root causes of issues that arise.  Narayan said that because the AI is embedded in the workflow, and ServiceNow uses a one platform, one architecture, one data model approach, there is the potential to use all these capabilities for automated resolution response. For example, if you’re at home and you spill coffee on your keyboard, it’s possible to speak to a virtual agent, have it understand your problem - and because the systems are integrated - have it know what keyboard you use, create a PO, and ship it to your home address without an agent being involved. Narayan said:  We have acquired technology and talent, but we take our time to integrate them into the platform natively. They are not standalone pieces, which means a single architecture, a single data model. Because if you don’t take the effort to put them together, it really kills your scalability in the long term.  A lot of AI could become AI for the sake of AI. One of the things we are super, super principled about is creating very purpose built AI. Using AI to drive actions, turn it into workflows, intelligent workflows, that deliver business value. There has to be a tangible ROI. 

The role of AI in a COVID-19 world As we’ve noted time and time again, the likely impact of COVID-19 on the future of the workplace and the nature of work is likely to be significant. It’s too soon to confidently forecast what that will mean in five to ten years, but Narayan was keen to highlight that there’s a new benefit to AI in the enterprise in the wake of the pandemic.  When companies think about implementing AI they often talk up the ability to boost productivity or gain efficiencies. However, the pandemic has meant that enterprises now need to be one thing more than ever - and that’s agile. Narayan believes that AI can help play a role in this. He said:  I think AI does two things for you. One, people think about it as making things more automated, more efficient, making people more productive, which increases your ROI. The second more nuanced point here is, it also makes enterprises more agile. If you think about COVID and people working from home, if you’re a CIO or a CSO and thinking about how to keep the productivity of your employees high, suddenly the types of problems that you’re needing to solve on a day to day basis changes a lot.  Zoom issues were not the biggest issues [pre-COVID]. You probably had one agent who knew how to handle Zoom or these typists of communication or VPN issues, but now they’re about 30-40% of what these companies are seeing. So if you’re not using AI or automated platforms, you have to think about how to scale your help desk. That takes a lot of time and resources, it’s expensive.  Whereas, if you’re AI enabled, it’s an automatable issue, you don’t have to have humans dealing with it. It can be a self service resolution or an automated resolution. The AI understands the patterns of the use cases coming in and then knows the recommended resolution. So it results in both a higher ROI and greater agility. 

It’s a compelling point and Narayan talks a lot of sense in terms of making AI usable in the workplace for the masses. But he’s also aware that given there is an endless list of possibilities for workflow enabled AI and that ServiceNow needs to be thoughtful about how it priorities its efforts. Narayan said that this will be the biggest challenge for ServiceNow’s AI strategy. He explained:  The challenge for us is that we have a large number of customers and lots of use cases. So how do we make sure we are disciplined in integrating these AI capabilities natively into the platform and understanding the high priority use cases? There are so many things to do, it’s like being a kid in a candy shop. I don’t want to go after every shiny thing.  What are the ones that are going to be very transformative for our customers? And how do we enable growth at scale for these use cases? I think that’s a challenge we need to be conscious of. We also need to bring our customers along on the journey. It’s not like we push things on the user, we do things in partnership with them. 

My take There’s a lot here that should please ServiceNow customers. It bodes well that the company is thinking in terms of integration with the platform and workflows, as well as focusing on use cases that deliver high reward. The outcomes will feel more tangible than if the AI work being done was locked away in a black box somewhere, only to be understand by data scientists. That being said, it’s also true that for all of this to work in practice and for it to be effective, companies need to get their own houses in order - and that means cleaning up all their data and systems, which will likely feed into the ServiceNow platform. Rewards won’t be gained in isolation, a holistic approach to being data-driven needs to be a priority.  

Image credit - Image sourced via Pixabay

Disclosure - ServiceNow is a diginomica premier partner at time of writing.

Tags
      
          COVID-19

Read more on:
IoT robotics and AI Machine intelligence and AI

Author: Derek du Preez

Date: 2020-12-02

URL: https://diginomica.com/servicenows-ai-chief-solving-last-mile-problem-enterprise-ai

diginomica.com