6 December 2019
The last two years have seen companies move their AI projects from the Proof of Concept stage and into production. Many of these are focused around a small number of processes or use cases, and often come from disparate sources.
PoCs have been successful, demonstrating the usefulness of AI in real-world settings. However, companies are beginning to see previously secondary considerations around AI coming to the fore – and in 2020, these will become critical to the progression of this technology in the enterprise. From customer experience to human-in-the-loop, businesses are now looking to how they can transform their offerings – not just on a single platform, but across the board.
Sanjay Srivastava is the Chief Digital Officer of Genpact, where he runs the company’s growing digital business, overseeing the Genpact Cora platform and all digital products and services. Ahead of Genpact’s participation in the AI Summit New York, we caught up with Sanjay to hear his five key predictions for what 2020 will bring – and his advice to businesses looking to make a true success of AI.
1: AI will increase the value of the human-in-the-loop
“There’s so much talk that AI will replace humans completely – I actually think it’s the complete opposite. My view is that AI will come to be an augmentation technology.
“Any time things get automated, the value of the humans that use it actually goes up. AI can make predictions, but once you’ve made a prediction you still need to make a judgment, and once you make a judgment you still need to act on it, and the latter two are attributes that humans have.
“We’re certainly going to see AI penetrate the enterprise but actually, as it does so, the value of the people that use it will exponentially increase.”
2: Customer experience design will be the true north for AI projects
“One of the things that will happen in 2020 is that experience, and experience-based design, becomes the true north for how transformation projects will take place in the enterprise.
“The way to think about AI is to first start with the business process you are looking at, and not to automate a big component of it, but re-imagine what that process looks like. Are you really trying to build a ride-sharing app on a mobile interface, or are you trying to change the experience of the rider from hailing, riding, expensing, and paying for a cab?
“The point is that you have to really rethink the experience that a client wants to go through – you have to journey-map their interactions and then you have to re-imagine what the new interaction should look like.
“What I find is that too many AI projects in the enterprise jump off of the burning platform of ‘we’re going to implement AI, let’s take this problem and just do it with AI’. That’s precisely the opposite of what should be happening. Instead, what’s the business problem, and who’s behind the vendor or the partner or the employee on the other side of it? What is the experience they are going through and what experience they would like? That’s what you need to try and capture.”
3: Digital ethics will enter the boardroom
“The ethics conversation has been going on for a while but it is coming to rest – at least based on what we’re seeing publicly around some of the regulatory work going on and the debate in the media around some of the large tech players. What has become fairly obvious is that the digital ethics game is not something restricted to large AI platform players. In fact, if you’re a company that works with AI or digital, it actually applies to you as much as anyone else.
“In 2020, I think that, for the first time, you’ll start seeing digital ethics become a sub-committee on boards and in board meetings. The same way that 20 years ago audit committees came into play and compensation committees came into play when we had all these excesses happening… Now it’s the norm for most responsible boards to have these committees. I think in the same way, ethics is going to take on the same level of importance. That’s the big change I see coming up.”
4: Getting AI off the ground needs to become quicker – and fast
“The biggest challenge of enterprise AI today is the time it takes to get to critical velocity is way too slow.
“What’s happening is this: if you take a horizontal AI platform and then you start taking data that you can normalize, manage, and get right and feed it through this horizontal platform, then you have to go and find the people who can reinforcement-train or label the data so you’re actually driving true machine learning. That’s a lot of work. Getting that to commercial roll-out takes a long time, at which point your IRR becomes very low and you don’t see success along the line.
“The way to change this is to start building with the bricks – these pre-trained AI accelerators that have been trained on data similar to yours thus you shorten the time to get it up and running.”
5: 2020 will see the emergence of transformation-as-a-service
“Nobody is deploying AI for the sake of it. What we are really trying to do is solve a business problem and transform a business service.
“So when you go back to transforming a business service, it’s clear there are many pieces to the pie. You’ve got this whole design piece which has to do with journey mapping so you can get that user base design. There’s process mining, which means figuring out how things are actually done today in a methodical fashion. That’s before you get into designing and thinking through the operating model which isn’t just the technology; it’s all the exception management work all the reinforcement learning work that still needs to happen.
“Then you have to do a lot of work on the implement side to get these AI things off the ground. You have to normalize the data, work out the data architecture, then you have to stand up a vertical stack for automation and then you have to get that right. So there’s a significant amount of effort involved in just building it which goes way beyond implementing a piece of software.
“Furthermore, you can’t just get an AI system up and running and wish everyone well and shut off the lights and move on. The problem is you’ll end up having data drift and the models will shift and the business processes will change and the interfaces will need to adapt in order to transform the whole experience.
“What I’m finding is that companies are saying it’s too difficult to bring these things together. They’re asking who they can turn to that isn’t just going to deliver an AI project, but instead is going to perform all the process mining, journey mapping, business process engineering, operating model design… then implement and deliver these AI capabilities, and finally is going to stick around to run it, govern it, and enhance it.
“That’s a big movement in the industry and I think things will change in that light – this emergence of transformation as a service and the increased realization that complexity based components are required to truly get use from an AI application.”
As told to Ciaran Daly
Sanjay Srivastava is the Chief Digital Officer of Genpact. Catch Sanjay and the Genpact team at the AI Summit New York, December 11-12