FEI Daily spoke with Rajeev Ronanki, Principal and Leader of Deloitte’s Cognitive Computing practice on the capabilities of cognitive technology, where to begin, and where the industry is headed.
FEI Daily: How has the use of artificial intelligence in business changed over the last decade?
Rajeev Ronanki: I would give you two perspectives on it.
One is we started looking at artificial intelligence about a dozen years ago, primarily to apply to our various functions: audit, tax and consulting. Portions of that are fairly labor intensive and, from our perspective, not particularly value add. We looked at applying AI to automate functions like data discovery, data collections and the predictive aspects of collecting the data and knowing what do we do with it and what needs to happen.
Rather than do it all manually, we started exploring how could we automate, how could we use technology so that we could be better, more accurate, higher quality, and in turn make it a more efficient process for our clients. A dozen years ago I would say we were really at the edge of an experimental research stage.
Over time we've evolved it to where now AI is a core or part and parcel of almost all of our functions. Robots to assist with labor intensive tasks, algorithms and machine learning that enhance the predictive nature of insights that we create for our clients, the entire gamut of things. We've really adopted AI as a core part of our business.
FEI Daily: What can cognitive computing do?
Ronanki: We tend to start with what are the core business issues that we've been trying to go after, whether that's the finance marketing, supply chain, what have you.
There are problems that existing technologies haven't solved. Either there's a lot of unstructured data that requires manual intervention to process, or natural language, or just a pace and scale of what's changing, the existing contemporaneous technologies haven't kept pace with it. That requires a different mindset.
What we tend to do is really identify all the points of highest value in any business process or function and then look at it and ask "Can the elements of cognitive technology solve that problem in a way that existing technologies can’t?"
If there is in fact that good intersection, that's where we recommend clients take a closer, harder look at it because the value will be inherent. It will be a compelling business case, and you're really focusing the technology on a specific problem as opposed to starting out with a horizontal question, "What can cognitive computing do?"
There's a lot of marketing hype out here. There's a lot of exaggeration of the capabilities of what it can and cannot do. The wrong path would be to just explore it without context. What we tend to do is the opposite, which is have a very specific context and a problem we're trying to solve and figure out the combination of technologies, so cognitive as well as regular descriptive technologies that would address it.
FEI Daily: How are businesses using cognitive computing today? How will they be using it in the future?
Ronanki: Today I think there are three main categories.
Number one is around process automation, which is much like at the turn of the last century, Ford automated the assembly line and therefore was able to create manufacturing processes at scale. Similarly, if you think about knowledge worker-based processes, the automation of that is being facilitated through cognitive technologies, whether that's processing a mortgage, or a car loan, or any of the processes of the healthcare arena like prescription fulfillment, or verifying a claim, or what have you. All these processes require a combination of structured data, unstructured data, and a set of rules, and a set of outcomes that are decided by humans. What we're finding is that you can train computing systems to mimic that, to process unstructured data, read and act on natural language, and then learn from the human decision-making processes in order to assimilate that and therefore augment human capacities to do that. I think that's probably more than half of the applications that are out there. In cognitive that's what's happening, which is intelligent automation.
The second is around engagement and consumer engagement. This notion of mass personalization has been around for some time, so rather than creating segments, and cohorts, and demographic groups, with cognitive technology you're able to truly create mass personalization at scale, so every single consumer in whatever the industry context, whether it's retail, or healthcare, or finance, or public services, every single person can be profiled for lack of a better word, and then a set of services could be created that are very much tailored. If you use Amazon Prime, it pretty much knows what you want to order, before you even order in. Think of that being applied at scale. Amazon is a great example of using that technology, and we don't even know it. Similar concepts apply to healthcare, banking, and any of the other industries. That's the second most frequent application, which is this notion of consumer engagement or, as we call it, cognitive engagement.
The third is around cognitive insights, which is finding the patterns and the insights from a combination of structured, unstructured, and voice, image, video types of data to find opportunities for either efficiencies in the enterprise or for growth.
As an example, if one of our clients essentially gets a continuous update of all the supply chain data across the world, and they also get an identified opportunity set around optimization. For example, China has got excessive inventory, or Malaysia has processing times that are lagging relative to the mean, and so on, and so forth. There's a continual self-optimizing set of insights that are generated. It's almost like a flood board for insights that gets generated every day with a specific set of business processes. That is a little bit more difficult to do, and it requires a lot of training and curation of data. It's something that's emerging. It's not widespread usage as yet, but I think that's something that's going to rapidly gain traction in the next few years, so it's getting more precise.
Read Part 2 here.