Скачать или смотреть Getting Started with Cognitive Business Operations

John details the benefits and challenges of Cognitive Business Operations (CBO) and how to get started.

Krista: What's the biggest benefit in using cognitive business operations?

John: The biggest advantage is robustness. Because it's a learning system and because you can present to it so many different situations, so many different variances on the state. So, for example, if you're doing case management where you're processing an insurance claim, every single claim is potentially different. With a cognitive system, you can adjust your training set to deal with all of these different situations then to be robust so that if something comes in that's slightly different, the system is still going to behave acceptably.

Krista: What's the biggest challenge when you're building these? You're talking about teaching and learning.

John: Right. So the biggest challenge with a cognitive system is that since it is a learning system, it has to have many, many examples to learn from, and that's really the biggest difference. So if you're starting from scratch with a brand new process where you have no historical information, your learning system is initially going to be not very good. Over time, it will get better, but really the quality of the solution is going to be based upon the number of examples that you can present to the system to learn from. So that is the biggest drawback.

Krista: So, what do you suggest for people who are starting from zero who don't have process data to fall back on? Or are you only suggesting that people who have long-running processes, that have been running for years, even dive into this world of cognitive business operations?

John: Somebody who's getting started with cognitive business operations, where they don't have a treasure trove of background information, where they don't have years of data that they can go in and create their learning sets from. I think the best approach is to go with that case management approach. Where you develop a solution where you actually have people who are able to look at the context of the situation and determine what the next best thing is to do. And then you're monitoring the choices that those people make, and that becomes your learning set. So, initially, it is a labor-intensive operation, but over time as the system learns, it can replace the human operators, learning from them.

Krista: Do you have a magic number? How many learning sets do we need?

John: The number of learning sets that you need. It's obviously dependent on the complexity of the process, but we're talking probably hundreds of instances minimum in order to determine what the best outcome will be. It's very, very, very similar to process mining. Process mining is where you go in, and you look at your background data in order to determine what your business processes are from the data. This is very similar. When you think of the amount of information that you would have to have in order to determine what your business processes are, using process mining. It's a very similar number to be able to provide enough data to the learning system, to be able to learn how to perform the process. Very similar number.

Krista: So, what are some ways that people can get started with cognitive business operations if we have long-running processes? How can we get to that information for the machines to learn?

John: One very good technique is to apply process mining to the data that you have gathered from those running processes. Process mining basically looks at data that has been gathered on the operations that are performed within your company. And it used analytical methods in order to extract from that, what were the business processes that were being carried out? What were the sequences of steps that were being carried out? If you have enough information for a process mining system, to be able to determine what your business processes are, then you also have enough information available so that you could begin training a cognitive system on how to perform those business processes.

Krista: Perfect. I think that's the video.

John: Excellent.