Hello. This is Professor Michael Rappa speaking with you from North Carolina State University in Raleigh, North Carolina, and I’m here to talk about my course, Managing the Digital Enterprise.
With the emergence of digital computing a half century ago, it was very common for us to think about computers and automation almost synonymously, which is to say that early computers were really seen as a replacement for or it enabled us to do things which previously were done manually, which is to say by humans, or largely by humans, with some other kind of mechanical devices, perhaps. And so the idea of computer automation or data automation was really quite prevalent. And I think this is something that we really need to revisit.
It’s somewhat ironic today that we spend most of our time in front of computers and working with computers. We don’t always think of computers as a replacement for manual labor or physical work, but in fact more and more of the things that we do certainly get incorporated up into a digital process that really is, in some ways, replacing manual effort. So I think it’s worth our time to revisit this question of how the digitalization of various kinds of business processes really translates into automation, and, really, where are there new and highly leveraged opportunities for automating things that we do either partially or entirely with human involvement.
Now, when we look across the landscape of digital automation, that is, when we look at efforts to use computers and, by extension, the web and the Internet as a platform for automating processes that previously required some degree of human involvement, you see a wide range of various kinds of activities. And the research and development along different fronts goes by different kinds of terminologies. And so in thinking about this wide variety of automation, I thought it was useful to add yet another piece of terminology to kind of capture this broad variation.
And so that’s where I came up with the decision to use digital automata as a kind of catchall or umbrella phrase for looking at a great variety of different kinds of technologies and research efforts which are essentially aimed at this sort of general idea of automating things for us, doing things for us that we might otherwise have to do ourselves, or at least have individuals involved in at some level. So the term digital automata is really one that you won’t necessarily see widely used but I think is really quite useful for capturing this great variety of activities in automation.
Now, one very common term that is brought up a lot, especially in connection with the Internet, is what are called intelligent agents. And I think this is a very important area. I think it’s just one, though, of many different approaches toward automating things that we might like to do. And somehow anything that involves intelligence catches people’s attention, and so almost from the start of the commercialization of the web people talked about intelligent agents, which are essentially software programs that carry out some set of operations on behalf of us as users and to some extent act independently or with some degree of independence. They may interact with other programs and in doing so act with some degree of understanding of what our goals and desires are in completing a particular task.
And this obviously covers a kind of wide variation of entities. Sometimes we talk about agents, other times we talk about bots or crawlers. The notion of shopping bots is something that was talked a lot about, the idea that as an individual I may have an interest in buying a particular item, and it would be nice to automate to some great extent that search process in finding out exactly the item I want to purchase at the price I wanted to purchase it for, and any other important criterion which I was wanting to employ in the search. But there are any number of different kinds of entities which sort of fall under this idea of an intelligent agent or program that really works autonomously on our behalf. And it’s not just shopping bots, but there are an array of efforts to automate processes that help us do things on the Internet which we might call agents.
And so, as we talked about earlier in the module on navigation, search engines today such as Google would not be possible without a high degree of automation in collecting the data from literally billions of web pages to store in its databases. And so web crawlers are a very common feature of what goes on in the background on the web today. Every major search engine has to employ such automation.
But there are also other kinds of agents. There are search agents that we use to look for things on the Internet. There are monitoring agents that might be used to look for changes in sites or content and send notification to the users. And I would expect that we would see just a plethora of kinds of automations of this sort whereby a software program is configured with the preferences of a user to carry out a specific task largely autonomously and may be more or less sophisticated in terms of its ability to learn user preferences over time to interact with other agents and so forth.
Another area of automation that I think is spreading quite rapidly and is a very significant, important part of commerce on the web today is what might be called a recommender system. And we see these employed now in various places. Perhaps a prime example would be a site we’ve talked about in the case study on Netflix. These are attempts to automate the process of understanding preferences of a user with respect to something such as, let’s say, their inclination toward movies, and to use those preferences not just of a single user but of literally millions of users to turn around and make recommendations to the user in terms of what they might like. So Netflix uses all kinds of data that it collects about its customers’ preferences in terms of the movies that they enjoy and do not enjoy and compiles this data together to turn around and recommend to the user, recommend to their customer, movies that they might enjoy which they haven’t seen. And we see this in other areas as well, and I think it’s one of the more valuable facets of such kinds of commerce on the web today, whereby user preferences are turned into recommendations of things that they might want to do or purchase.
I think the recommender systems we see today vary in sophistication, perhaps Netflix being on the leading edge. But if you look at sites such as Amazon is one example, you see a kind of range of technologies that essentially try to correlate the things that you do and the preferences that you express with items that you might want to purchase. There’s little doubt we’re going to see the use of recommender systems continue to spread around the Internet, and I fully expect that we’re going to see just tremendous advances over time in their sophistication and accuracy and ability to do things which are useful for us.
Now, one area of automation that has captured our attention for decades and decades and decades really has to do more with the representation of humans themselves in doing things for us. So whether it’s a physical manifestation like robotics, we also see in the digital world a tremendous amount of interest and creativity going into the human representation within a digital environment. So whether these are things we sometimes call chatter bots or avatars, sort of virtual assistants that do things for us with a kind of human touch, whether it’s read our email or news or handle our personal calendars or service reminders, that we are essentially interacting with something which has some humanoid characteristics in either a kind of physical representation or other human characteristics such as voice.
And, of course, anyone who is involved in the gaming side of the digital world knows that there have just been amazing advances in terms of the representation of human activities within the digital environment. And certainly there’s just a lot of excitement there, and I think we’re going to begin to see the importation of some of the things that are going on in the gaming world across a wider range of activities that basically help us interact in the digital world in ways that are perhaps more kind of user friendly from a kind of human interaction perspective.
Now, I think we’ve got some ways to go, and some of the examples one might point to may seem not so impressive at this particular stage in development, but certainly we can expect to see advances in the future. And of course we really shouldn’t underestimate the degree of difficulty there is involved in trying to not only represent humans in a kind of digital environment but also to do human things, whether it’s understanding speech and being able to respond or incorporating other kinds of higher levels of intelligence. These are long-term research agendas that I think we are seeing progress on, but we still have a ways to go.
I think it’s really important not to get too focused overly on the representation of humans or human action and the kind of glitzy side of this automation. I think the most valuable, the most important, the most significant advances in digital automation really have to do with much more mundane things that don’t end up on the front pages of newspapers but yet end up saving businesses and saving organizations an enormous amount of cost and effort involved. And so if we really turn our attention to some of the simpler forms of automation which nonetheless have a high degree of leverage to them, we can really see where things are heading.
Let’s take the example of customer service. Customer service is part of every organization, and it can be a very expensive component. That is, fielding questions and issues coming from our customers and resolving them in an effective way. One of the things which we came to understand fairly quickly with the emergence of the commercial web was that if our customers were given an opportunity to find answers to the questions that they had, they would largely do that on their own. So if on a web site we could provide the means for searching through and finding the answers to a particular question, that generally speaking customers were inclined to take advantage of that opportunity, and maybe even for many people to take advantage of that kind of self-service approach to finding answers than picking up the telephone and calling a customer service center and speaking with an actual person.
Of course, this inclination may vary depending on the customer and the customer’s age and their comfortableness with using the Internet, but I think if you turn to just about any major company nowadays and you look at what they’re trying to do over their web site, it’s not uncommon to find a significant component dealing with customer service. And so one can think about, then, the web as providing an opportunity to partially automate that process. That is, taking the customer service rep out of the equation so that the person with the question, the customer with the question, can now turn to the web and find answers on their own.
Of course, anybody who has done this knows that depending on how a site functions, that can be easier or more difficult, and so there’s a certain degree of intelligence incorporated into how do I automate that side of providing customer service over the web. How do I make it simple, or as simple as possible, for customers with questions to come to my web site and find their answers quickly? So if you begin to examine that problem and you understand, well, we need a database with a range of answers to questions that my customers may have, and then I need a fairly robust query mechanism that would enable a customer to ask a question more or less in natural language.
And, of course, any particular problem may have a number of different expressions in terms of how the question is asked and when a question is asked that it yields the appropriate answer at the top spot or the closest to the top as it possibly can get. And so when you think about that aspect of it, people querying a database without a lot of sophisticated knowledge about how to pose questions in a structured manner, how do we deal with real live customers asking questions in the only way that they may know how and yielding them quickly with the right answers, that’s a pretty challenging task. And I think a lot of effort is going into that kind of automation of the customer service rep function, and especially in the area of deciphering and understanding natural language queries so that rather than the customer going through a kind of artificial step-by-step process that might narrow down the appropriate answer to their question, instead they can simply come, ask a question the best way they know how and ultimately yield them the correct answer.
Now, some of you are probably thinking, oh, customer service, it’s kind of a mundane area of the business world. And yet when we look at customer service, it really is a kind of critical function in any successful business. Number one, you want to have satisfied customers, and customers are going to have questions about the products that they’ve purchased. And so whether beforehand or post-purchase, it’s really quite natural for there to be questions. And providing them with quick and easy methods for finding answers to those questions is going to yield more or less satisfied customers.
The other side of that is that when we use people on the customer service side of the equation, then we’re really talking about, number one, what is potentially a significant expense in terms of training, in terms of retention, and to the extent that they’re handling questions which might be simple to handle in an automated mode, we could really save ourselves as a business a lot of money if we can do this in an automated fashion. I’m not saying that it’s necessarily possible to eliminate the human customer service representative, but we want those real live people to be dealing with the most difficult kinds of questions and issues that are raised by our customers and not waste their time with the trivial ones.
Anyone who’s worked in customer service knows that it can be a really grueling and difficult job, and when you talk to companies, particularly in businesses where customer service reps are dealing usually with very difficult customer queries, you know that it can have an enormous amount of turnover, and that just adds more and more to the cost. The more people that we have to recruit and train, and to the extent that they only last 6 or 12 months or 24 months on the job, it just costs us a lot of money as a business, and so what we can do to automate that process is obviously going to be an important goal for businesses.
So if you spend some time thinking about customer service and what can we do to increasingly automate that process as we interact with our customers in an online world, then all kinds of creative ideas begin to emerge that say can we anticipate customer service queries. Can we help customers before they have problems? Can we use the fact that we are receiving all of these queries from our customers now in a digital environment to really understand the kinds of problems that our customers might be confronted with and solve those problems more quickly or easily as to prevent them from happening in the future? All kinds of ways that, again, because you’re capturing all of this information as part of a data stream, we can kind of turn around, learn from it, and then incorporate that learning back into the digital process. And so I’m very hopeful. I’m very excited about where this is heading in the future and the potential for just more and more creative effort going into how do we provide a better experience for our customers through this kind of automation.
Well, I think we can talk about other areas of business processes and beyond customer service, although I think customer services is a significant area and one that’s easy for all of us to understand. What I’d like to do is kind of move away from thinking about automation purely as a beneficial aspect of the digital world and turn our attention to other kinds of automation which sometimes are very problematic for us.
When we turn our attention to some of the more problematic areas of the digital world today, what we see is really the clever exploitation of automation. So if we look at, for example, spam, the torrent of irrelevant and annoying emails that we may receive in any given day, this is largely done through automated processes. I mean, one doesn’t sort of sit at a computer and send out millions of email messages to folks with a lot of human intervention. This is all done with automated systems which send a barrage of messages across the Internet.
Again, also, we look at worms and viruses. These are automated tools. These are programs that are acting at their creator’s directive, and they are, as we’ve seen over the last decade, becoming more and more sophisticated in their ability to mutate and to find their way across the Internet and into systems that are not as secure as we might want them to be.
Automation is both a great opportunity for us in terms of saving us effort in doing the kinds of important things that we want to be doing as businesses, as consumers, and the like, but it also represents the most vexing and dangerous part of the digital world as well, in the sense that one can automate an enormous amount of destruction and annoyance. And dealing with that side of it is really quite a challenge for us and will continue to be a challenge into the future.
I think that many people have a kind of misconception of sort of that kind of hacker sitting in front of their computer and trying to find their way into some remote system. Maybe this is the result of too much of Hollywood in creating a kind of mythology around hackers, but when we look at what happens today on the Internet, what we see is that the efforts to compromise various networks and computers that are connected to them, that these are for all intents and purposes automated attacks. These are processes that are set in motion that go looking for vulnerabilities from system to system to system. And it’s only when those vulnerabilities are found that someone may step in and seek to take advantage of it.
And so automation, perhaps some of the more interesting and important creative effort today is unfortunately going on in automating processes that lead to the compromise or destruction of data or other kinds of harmful things to systems connected to the network. This is a subject that I’ll have more to talk about when we come to the topic of security later in the course.
Well, I think one of the more fruitful areas to spend some of your time thinking about is precisely this question of how do I incorporate more and more sophisticated automation into our business processes. I see this very much as a kind of stepping stone process. As we look at a process and begin to automate those things that we’re able to most easily, one can really kind of see a progression of moving step by step by step so that more and more of the process becomes automated over time. I think there are strategies to attack that problem in terms of looking at where costly repetitive actions that are easy to automate provide perhaps an early opportunity in allowing some critical thoughtfulness to understanding what part of this process is really the most difficult, the most heterogeneous in terms of really needing the intelligence of our people as an organization to be intimately involved in it.
If we come back to the customer service example and we think about the kinds – if we looked at the distribution of questions our customers had about our product in any given period of time, we would probably see a situation where there’s a great degree of redundancy in terms of similar questions, and those questions may be fairly basic and easy to answer. But there’s a part of that distribution which gets increasingly difficult, whereby the questions are idiosyncratic, where all of a sudden the cost of automating the ability to, say, deal with that part of the distribution becomes much greater and you begin to see where having people involved, having a really human customer service rep there to handle those kinds of questions, becomes more cost effective. They’re dealing with the difficult problems that really require some human intelligence and ingenuity to find what the appropriate resolution or solution is. So I think that this is kind of part of a kind of constant evolution to use advances in technology to incorporate and automate more and more of the things that we can and leave the real difficult stuff to us humans.
Let me conclude the conversation with an encouragement for you to look at opportunities to automate processes as you move into a digital environment to provide either more value to your customers or to turn the attention of your people to more productive activities through automation.
This is Professor Michael Rappa, until next time, wishing you all the best with your studies.
(Music)
[End of Audio]
Unedited transcript of audio podcast produced on October 18, 2005.
Audio source file: https://digitalenterprise.org/podcasts/automata.mp3
Michael Rappa is the Alan T. Dickson Distinguished University Professor of Technology Management at North Carolina State University.
For more information, please visit: digitalenterprise.org
Copyright 2006 Michael Rappa. All rights reserved. Please do not reproduced, distribute or quote without written permission of the author.