Panel Discussion, Taken from our Ascot CIO Event, October 12th 2017
Professor Sharm Manwani, Henley Business School (Moderator)
Bhavn Khaira, Head of Group IM, Element Six
Ela Ganesan, VP, Citigroup
Durvesh Ganveer, Chief Architect, NTT Data Services
Neil Ward-Dutton, Research Director, MWD Advisors
Bhavn Khaira, Head of Group IM, Element Six-
Element Six makes industrial diamonds and there is huge potential for AI to play an active role in terms of increasing operational efficiencies, increasing automation and digitisation, using that data to improve the shop floor, managing yield and supply chain a lot better and therefore improving the customer experience.
Ela Ganesan, VP, Citigroup-
AI has been here for a long time so what has changed from our perspective is that there is the ability to process more data at a given time and the software is playing a big role. So we are now trying to pilot all these cognitive services and all the stuff in Siri, but that said, I do consult with some of the cloud setup as well. So things are different there, so we are trying to see how AI can serve a grand purpose but at this point in time I would say we are doing a number of pilots but those are at very early stages at the moment.
Durvesh Ganveer, Chief Architect, NTT Data Services-
From a services perspective, the most important question is ‘when we say AI, do we all mean the same thing?’ We talk about, augmented, assisted support to engineers, we talk about automation and we talk about virtual assistance.
So I question whether we are at an age where AI is there in the industry or not. We have data, we have analysis of it, but we just use a rear view perspective on the solutions that we deploy.
Neil Ward-Dutton, Research Director, MWD Advisors-
My particular focus is on automation technologies. So for the last 10 years or so I have been looking at how automation technologies of different kinds have been developing and how they are being used in a whole variety of industries.
So clearly AI has a big potential impact on the way that people do tasks, tasks get coordinated and decisions get made and my perspective really is based off the technology we research, case study research and benchmarking, understanding how organisations are trying to get value out of this and what’s the difference between an organisation that gets value and one that doesn’t.
From our perspective, having worked with a number of manufacturers as well and given the context of the industry, it’s not been a straightforward journey. We’ve already invested in a lot of legacy infrastructure for example and manufacturing is very capital intensive and there is also resistance definitely from an executive level to say ‘what does this actually mean for us as an organisation?’
If you look at the .com boom when everyone invested quite heavily into exploring what opportunities were there, it required a lot of resources and everyone jumped on the bandwagon. What I have definitely seen from a manufacturing perspective is that people are reluctant to put all their eggs in one basket and not do massive capital expenditure into what, for example, Industry 4.0 is the equivalent of in manufacturing.
But what we are looking at is more pilot projects and a very agile approach to say ‘we have a pot of money, we will try pilots to see exactly what works for our organisation and what doesn’t in a way that doesn’t take away from the daily production but in a way that we can see measurable improvement and then agree which is best for us.’
What works for our organisation wouldn’t necessarily be the same for another and I’ve seen that across a number of manufacturers.
For us it is at a very early stage because financial services tend to pick up these technologies very early, but that said we don’t put anything into production until we know it is going to serve the purpose or resolve the problem.
We started everything as an extension to the existing capabilities, so we aren’t saying we are going to use AI for everything but what we are doing is putting the extension to the existing capabilities and saying for example ‘what can we add to the analytics that we are doing now? Is there anyway AI can come in and do things better?’
Then we are trying to see some of the customer experience for the consumer banking as well where we are trying to see where tools like chatboard are useful in chasing the customer experience.
It’s very important that we find a balance with the amount of data and the quality of the data. So when we try and add machine learning, it’s very important that we have the right quality of data and the ability to feed those datas to try and get value out of them.
There are companies like Microsoft and IBM that have their own cognitive services but we are really careful about picking those services because we need to find a purpose because we are trying to solve an industry problem.
In terms of the whole concept of AI, we have gone down the path of defining what exactly we intend to deliver under that umbrella. We have a rich library of deep analytical insights that have been developed from information sets collected through our services engagements such as service desk analytics, end user analytics, customer experience analytics and enriching the user experience.
Saying that, they are data sets or insights that we have quantified into robots. We have about 2,200 robots in production in the US in the healthcare sector but I would still not class this as AI because every robot that we have programmed is a one trick pony. It does what it is meant to do and that’s it.
The engine that we have built around it that helps us orchestrate those robots is again just a cognitive engine that knows how to deal with these 2,200 robots. Put in a question that is slightly different from the most complex one that we have and it has to go to a human and if it cannot address that scenario, you do not have AI. You only have a very rich library of robots.
So what we have done is we have built this ecosystem; a set of dashboards that tell us which robots are in production, how many times it has been used, what patterns are coming out etc. Leveraging this knowledge with the human intelligence, we can continuously enrich this library of robots.
We have put them in production, we know the consequences of certain actions taken by robots, so we know consciously know what actions a robot should and shouldn’t take and we have introduced the appropriate escalation point at that place.
We have integrated voice into actions with end users through virtual assistance, which is also not AI as they are still running the scripts that they have been programmed to address. That integration point with the cognitive engine and the user analytics that we capture, we are continuously creating use cases for additional robots.
And we have bots that can create bots because the whole pattern is based on reusable components. So we look at it as robots that are robots and we are still a few years away from AI.
The principal way that we are moving things forward is through discovery workshops with clients, because it is still very early days and we find a lot of people in our community are not really sure what this is or how it can apply. So we are trying to help people through that.
The challenge is that AI is not a thing that you stick in the corner, you plug in and magic happens. If you are going to think about AI at all, think about a whole cluster of different kinds of services and APIs you might hook into that all do different things for you. Fundamentally, as far as we can tell, there are three layers within this clustering:
At the top is Interaction- there is a whole cluster of things happening in technology that are to do with how we interact with machines such as Visual Recognition, Classification, Natural Language Processing and Generation, Speech Analytics and Sentiment Analysis. Fundamentally, they enable computers to sense and respond to things in a more human-like way.
Secondly is Insights- tools and technologies that enable machines to make more in the moment, predictive recommendations about what’s gonna work or not work in a particular situation.
Finally at the bottom is Integration- as mentioned, things like Robotic Process Automation and Bots.
All these things are happening at once and it’s very easy to think of them all as the same thing but actually they are very different. They do different jobs and they deliver different kinds of value in different situations.
So what I spend a lot of my time doing is helping organisations map those technologies onto their own operating and capability models and figuring out where and when they will get value from them and we end up with a very diverse set of use cases and benefits.
I definitely agree with Neil; it’s down to identifying where do we think we are going to get the biggest opportunity from an AI perspective, but also the definition is key which Durvesh raised.
If I look at it from a purely manufacturing context and an Element Six perspective, the machines themselves are stand alone machines and we have to make them more intelligent by connecting them and providing a way of connecting them through sensor technology, then collating data from that to start being able to get smarter about what these machines are actually doing.
When I went to the Siemens factory in Nuremberg, they had so much more digitalisation. Their key driver was around operational efficiencies and then you saw the level of automation that was brought in, they had obviously looked at the consistent process flows and looked at real opportunities where automation and smart-manufacturing was playing a key role in that.
I haven’t seen anywhere where we have used ‘true AI’ yet in the manufacturing sector. Definitely automation and digitalisation, but I definitely believe that the approach that Neil talked about is where the majority of the industry has gone, from what I’ve seen.
I think from the financial services perspective, we don’t put anything into production unless we know that it is delivering value to the business problem we are trying to solve.
So the two use cases I have seen in my experience would surmise that there is a need for improving the customer experience. It depends on multiple factors like the data and interaction you have and the perimeters for that interaction and how that relates to your reply to the customer.
It takes a long time to train and retrain and get them to self learn as well. We tend to say that financial services are picking up AI learning at high speed but we are into real time yet.
There is this new saying that ‘data is the new oil’. My concern with that is that if that is fundamentally the case, we collectively are the prehistoric animals that have been squashed to make the oil. You think about where all this AI technology has come from, what is driving it? Google, Amazon, Facebook etc. are training their algorithms on our data.
So one of things that concerns me is that we get an amazing innovation out of this, but the way it is coming about, because of ad-supported models, is that fundamentally it is our private data that is training a lot of these sets.
So it’s really about privacy and how we manage data within our organisations so that we can respect people’s privacy and deal with the data responsibly.
On the same topic, with data being the oil, you introduce a new letter S for soil and that generates newer insights. So the concern I have is around controls.
What is the level of control that you are aware of and need to put in? And then what are the level of controls that you don’t know today that you need to put in that are likely to expose your data to the point that private information is being revealed?
You have missed out some details, please try again.