Q&A: Jonathan Crane, IPsoft (Part 2)
We continue our wide-ranging discussion with Jonathan Crane, Chief Commercial Officer at IPsoft, about automation, cognitive capabilities, his organisation’s groundbreaking Amelia product, and the impact of all this upon today’s outsourcing space and the broader business environment… To read Part 1 of this interview, click here.
Outsource: Let’s take a look at Amelia, then. At last year’s Aecus Innovation Showcase event one of your colleagues described the way in which Amelia has been designed as being intended to mimic the structure of the human brain – can you shine a little more light on that?
JC: Yes, that analogy holds. What you’re looking at is Amelia learns as a child would learn, as you or I learnt. There is an ability to input into our brain: I can do that by observing, I can do that by reading, I can do that in many ways. I take that information into my mind – and, looking at Amelia, she will take that into her mind and she will populate her brain. We’re talking about neural ontology.
Let’s say we give her your background: for the first time we’re going to introduce you to her and we give her all the specifics about your background and accomplishments. She would then map that, in her brain, and she would be prepared to answer questions or enquiries relevant to you based on how she has learnt what it is you have shared with her about you. So she would be able to answer questions like “Where were you born? When were you born? What were previous writing accomplishments you may have?” Whatever was in your bio, she’d be able to do that.
Now jump forward and think about business scenarios Amelia could learn: she could know claims processes in an insurance business; she could know how to manage various vendor enquiries if you were a particular supplier. There are a whole host of things she could do. The starting point is providing her with knowledge either by observation or direct input. In addition, she gains this knowledge across a horizontal environment, meaning if you teach her how to do claims processing in one insurance company there is then a baseline of intelligence relative to that processing of claims, and you can then augment that – or delete – or enhance – or add compliance – or whatever you want to do to that baseline. So you get horizontal benefit from what we’ll call the knowledge library: what she’s learnt about particular business processes in a particular vertical industry.
So that’s how she learns: she starts as a child and becomes a PhD, in a faster time than a human.
O: Where in terms of industry verticals have you seen the biggest take-up thus far, in terms of those organisations trialling Amelia?
JC: Clearly one of the first and biggest is the financial services market, considering the complexity and breadth of the kind of services they provide – everything from wealth management to monitoring simple things like cheques and all the related authorisations. You could, with that example, have her scan a cheque and know immediately whether this cheque is or isn’t applicable. You could do this on order entry or any number of other client opportunities. You could do this as a service desk: we’re doing this today in a customer service environment, taking in a substantial amount of volume. Volume has forced this organisation to hire more and more people because the average time of complexity for solution is about 18 minutes.
With the Amelia trial we’re down to around four minutes to resolution, and the answer time is almost negligible. Computer answer capability is immediate versus around 55 seconds of hold time using traditional approaches before you get anybody even answering the call. You can see the applicability in a service centre, where you are supporting vendors, partners, end-user customers and so on.
Frankly we’re just in the early stages of trying to think about how to apply Amelia, and cognitive technologies, into business processes today – which says there’s a lot of opportunity left. McKinsey say that cognitive based automation of today’s knowledge work is a market that’s defined by trillions, not billions – and therefore it’s got a lot of people’s interest!
O: Where does this fit into the broader cognitive landscape? There are a few companies making big claims about the tech that they’ve developed… Do you see the landscape that’s being discussed at the moment as actually representing the technology that’s out there and what can actually be achieved with it?
JC: If you look at IBM’s entry into the market, it opened up the opportunity – but more in terms of what we’ll call statistical analysis. A massive compute. Accumulating data and knowledge that would eclipse a human’s ability to absorb and then act upon without being overwhelmed by data. This is a skill set that IBM clearly possesses; they have infrastructure and software that can support that well. So they take on massive-type problems, do some statistical analysis, and are able to provide numerous solutions based on the data being assessed – are able to tell you what’s a likely outcome, and that kind of environment. That’s going to be very powerful, the ability to manage data – and the buzzphrase ‘big data’ – and manage in a complex fashion lots of disparate data and bring it to some kind of solution. I think IBM has pioneered that and continues to play in that arena.
The other side of it is the Googles of this world, who are having great success in some of their projects, most notably the car example; there it’s a phenomenal opportunity to look at if we need to drive or not. Is there something we can change about our behaviour where cars don’t need to be driven by us – in fact they’d be better driven by computers – and oh, by the way, think of all the knowledge we’ll gain, of all the input a computer will be doing driving a car versus me. I can be distracted by my cellphone; I’m listening to some tunes, or a news broadcast; I’m not really paying attention to a lot of things that I could be when I’m out on the road. So, again, accumulation of data is going to be huge for Google within a specific application.
Where we really see this going is everything from – think about machine-to-machine. Think about the ability to do the work of a dispatcher, who would normally dispatch – let’s use the example of people to soda machines. That dispatcher has the ability normally to try to look at and manage routes as best they can, but frankly it’s hit or miss depending on how much knowledge you have about what machines need to be replenished faster than others. Now suppose you knew everything about those machines, and that every one of the sodas that are in there are consumed at a certain rate. Add to that, a predictability analysis that’s already been conducted; you know that it’s going to rain today and that as a result not as many sodas will be consumed, versus it’s 90 degree, hot and sunny.
These are all things that can be done. Combine this data with Amelia at the front end so that as she interfaces with the drivers she can pull all the relevant datapoints when they ask for directions and which machines to go to, or what product they should have on their trucks to stock.
Or you could put Amelia in at the front end in the health industry: you come in, sit down, get your blood pressure and other vitals measured. Then Amelia converses with you and asks what the problem is; you tell her; she does some quick analysis, recognises the symptoms and factors in the knowledge that the 22 people who preceded you into that kiosk had similar symptoms which indicates there is a flu outbreak. In fact, it’s quite possible that as well as diagnosing the symptom Amelia could also prescribe a drug.
The possibilities really are broad. As long as Amelia is connected to a service centre and is given permission to handle routine problems she will be able to create efficiencies. Think of common questions posed to IT service desks: such as “I forgot my password” or “I need my password reset”. All these recurring questions burden helpdesks and could easily be taken and put into a process handled by a virtual agent.
To read the concluding part of this article, click here.