This is a podcast on the practical ramifications of artificial intelligence project rollouts in business. It was sent to me by Nick Cordrey. You can download the podcast here

I do not own the rights to this podcast. I loved it so I had it transcribed. I have posted it in the hope it will help others.

Good evening, ladies and gentlemen. Welcome to the LSE for tonight’s event, a Department of Management and Public Lectures.

  • My name is Edgar Whitley and I’m an associate professor in Information Systems in the Department of Management here at LSC.
  • And it’s my great pleasure to introduce Prof. Leslie Willcocks and Prof. Mary Lacity.
  • They have an unprecedented access to organizations that are the early adopters of what they call robotic processing or automation, and in their latest book, they capture years worth of learning about embedding these disruptive technologies into the modern workplace – seeing this as a blueprint for the future of other organizations moving forward.
  • Tonight they’ll provide a balanced, academically-informed and compelling view of the many benefits as well as the practicalities of managing these kinds of feature and technologies in the workplace.

Good evening, ladies and gentlemen. Mary and I are delighted to talk about our book this evening with you. Service Automation Robots and the Future of the Work.

What we have is an agenda which is I have to admit quite a mixed bag.

  • There’s academic nitpicking with some serious consequences despite the rhetoric about smart and intelligent machines.
  • Some examples of machines and people getting it wrong.
  • Some cold water on the idea that people will not matter in a highly technologized future.
  • Hard won evidence on actual automation deployments in major companies.
  • A brief review of the impactful technologies.
  • A sceptical view of the future prognosis and some projections of our own.

By last week, even the Financial Times was running a four-day special series on robots, friend or foe. This feature is now a familiar litany of story lines, driver-less cars, an AI writing program called EMMA, Google’s — defeating a champion of the ancient game of go, a spring of technologies that can already do physical and cognitive tasks better than humans, the news of a US investment boom in artificial intelligence, together with commentaries from optimists and the pessimists alike.

  • As a result of this development, our study is framed in this much bigger context and looks at the timeline from 2016 to 2025.
  • Frankly, when you look at the bigger picture, it is not easy to kick your way through the media representations of the debate around automation robots and the future of work.
  • There are many detailed service studies of course such as by Richard and Daniel Susskind’s The Future of the Profession, which they delivered on this very stage several months ago, and Martin Ford’s The Rise of the Robots.
  • These provide much detail in the tasks that could be automated in multiple sectors and across most occupations in the relatively near future.
  • Our own work suggests a more nuance and complex narrative than the headlines showed to us on a daily basis. And less pessimistic conclusions than Ford and the Susskinds arrive at.
  • Our book also suggest a future shaped by human will and imagination and ability to absorb change as much as by any specific technology whoever heavily invested in.

I would make another point especially about the more macro studies on the technology and future job numbers. Reading has made us increasingly skeptical. The problem with nearly all the studies is that they are projections going forward with not necessarily good data sets, often carrying questionable, even tacit assumptions and few make their methodology transparent.

Even if you take the best of the studies, the limitations start to become clear quite quickly.

So let’s take one of the best of the studies and we will begin to see what I mean.

  • Frey & Osborne 2013. Probably the most rigorous and admirable of the studies that we will talk about today. Certainly one of the most quoted. And their finding, looking at 2010 data for 702 occupations in the USA, here they found as this graph shows on its right side that 47% of occupations were under high risk of being computerized.
  • The figure to the UK in their latest study is 35%.
  • However, self-admittedly, the research is to not try to specify the speed of technology developments nor a time period for the loss of jobs. The best they come up with is perhaps over the next decade or two. Nor did they attempt to predict the actual number of jobs lost.
  • Referred and limitations we would point to but remember this is one of the best studies.
  • Personally the study like many others in this area, there’s no analysis of jobs likely to be created by changes in the work technology. More on that later.
  • Secondly, it focuses on jobs and occupations but not on activities, nor the amount of work that needs to be done
  • Thirdly, the study largely fractures out the key bottlenecks of how commercially feasible, viable, organizationally-adoptable the emerging technologies are, i.e. the long roads of the fusion of animation dilemma is ignored.

But they do bring one thing to bear here that’s very interesting. They talked about three engineering bottlenecks to computerization.

  • The bottlenecks are written up there and essentially they represent things that are not being engineered very easily and that cannot be automated.
  • But in our view, these bottlenecks are understated, and the three concepts are not adequate for fully describing the multiple, valuable human qualities that will continue to apply at work. You can make your own list, but as a starter, think of leadership, adaptability, self-awareness, critical thinking, tacit-knowing, conflict resolution, managing emotion, imagination, composite skill use, ethical judgment.
  • I did this recently with a group of IT people and we spent 40 minutes drawing up such a list. Jeff Cullivan wrote a recent book called Humans are Underrated, and Tom Davenport and Julia Kirby have a fourth book coming one entitled Only Humans Need Apply. Both are salutary reading on this subject.

But I want to drive on a little bit down this point and really reinforce why humans continue to matter in the next 10 years.

Mary :

<I’ll cover>

  • ‘How are businesses actually adopting this new breed of service automation technology?’.
  • And then at a more global level ‘What are the implications for global employment and the future of all of our work?’

Businesses and how are businesses are adopting this new breed of service technology.

Leslie and I decided we need to go look at early adopters, organizations who have already embraced this new breed of service automation technologies to understand what kind of outcomes are they getting, how is it affecting their employees, how is it affecting their customers.

  • So we did so far 13 case studies and you’ll see some of the logos of the companies we have up there.
  • One is you’re gonna notice that they are cross industries. So this new breed of service automation technologies. They’re not industry specific. They’re not process specific. They’re general purpose tools that are designed to automate a whole series of different business services.
  • The second point about our survey, I mean our case studies, is that the sample is completely biased. See, I told you I’d be honest. It’s biased because we only studied success stories.
  • So all of these companies have deployed this new breed of service automation technologies. They’re getting great business value delivered for their shareholders. Their employees actually embraced the technology instead of being intimidated and worried that it would take over their jobs and it ended up with better customer services.
  • I’ve used the term the New Breed of Service Automation Technology several times. So now I wanna put some meat on that. What do we mean by that?
  • We’re talking about software.
  • Operating in this space are literally hundreds of companies and there’s a big Tower of Babel out there of the different terms you might have heard of from cognitive intelligence, automation or cognitive automation, robotic process automation.
  • All of the tools that are designed to automate services with these characteristics we’re going to call it Robotic Process Automation.
  • And these tools are designed for structured data, rules based processes in the form of if-then-else statements, and have deterministic outcomes.
  • So all of those tools have to do with robotic process automation.
  • Now those of you in the audience who should be thinking ‘Haven’t we been automating structured data, rules based process for 30-40 years? What is different about the new breed?’
  • We’ll talk about some of the differences but the main one is the ease of use.

It used to be then when we were automating services with these characteristics, we needed to go to an information technology department and have a whole team of IT professionals do the automation on behalf of the business.

  • Now you take people who understand the business process and they don’t program the software. They configure the software with a GUI dragon drop interface and they are the ones now doing the automating.
  • Now we’re gonna walk over here which is more in … probably what you’re aware of in the popular media, the realm of Cognitive Automation.
  • Here we’re gonna put all of the tools that are designed for data that is more unstructured, largely textural data, data that you communicate with these products through natural language interfaces, so either through voice or through typing and they respond to you in a natural language that the processes underneath are more inference based rather than in a form of if-then-else rules.
  • And the outcomes are largely probabilistic rather than deterministic. The big Uber machine in this space is IBM’s Watson. And so we’ll get a little bit more into the Bell Weather Event of IBM’s Watson winning Jeopardy, and then we’ll tell you where we’re seeing organization in their cognitive automation deployments today.
  • So we circle the realm of RPA because of the 16 case studies that we’ve done so far. 14 of them adopted Robotic Process Automation. And only 2 of them are starting to experiment with a more cognitive automation.
  • And there’s a good reason for that. RPA technologies are ready to go. Companies can go from zero deployment to implementation in a couple of months.
  • The upfront investment to deploy these tools is much lower than on the Cognitive Automation scale. So we’re calling RPA today’s technology and we’re calling Cognitive Automation more of tomorrow’s technology from a business perspective.
  • This is one reason why I’m going to put this survey up here ’cause I mentioned we did a survey in 2015 at the Outsourcing World Summit. One year later, we replicated the survey and the uptake of the adoption of RPA was astonishing to us.
  • So we saw a huge uptake in the adoption of RPA technologies in one year.
  • We asked the same group ‘Where are you in your adoption of Cognitive Automation tools?’ And we still see quite low levels of adoption.
  • So it makes sense that we’re gonna talk a lot more tonight about RPA than Cognitive Automation but before we focus on RPA, I wanna bring you back 5 years ago. Okay.

Can you believe that it’s been 5 years since IBM’s Watson won Jeopardy?

let’s take under the hood of IBM’s Watson and what’s actually happening here.

First, when the question gets asked, there might be up to 100 different natural language parsing algorithms going on to just interpret this question to begin with.

Then, Watson starts generating hypothesis what they think the answer could possibly be and it starts going against the database of 200 million text documents. All the contents of Wikipedia, the Bible, literary works, movie databases, and starts waiting each piece of evidence until it comes up with four solutions and the probability that it thinks it’s correct.

So I want you to be amazed and awed by Cognitive Automation. But that story I just told you, awesome means that when organizations adopt these more complex tools, it’s a much bigger investment.

So whereas RPA can be deployed in an organization in 3 months, we’re seeing organizations who are adopting big cognitive automation tools taking a year, 2 years, 3 years, even more using Cognitive Automation.

So it’s a bigger investment but the applications are gonna be wow.

Switching to RPA

Now, we’re going to switch over to Robotic Process Automation. That’s today’s technology.

We’re seeing huge widespread adoption in businesses, and I wanna tell you a little bit more about Robotic Process Automation.

So I mentioned that one characteristic of these tools is that they’re so easy to use, that business people can do the configuring of the software to automate a human task rather than having to go to an IT department.

The way RPA works is it gets a log-on ID and a password just like a human being does, and it’s perfectly suited for those structured, the structured data, rules-based processing tasks that a lot of humans do.

So we tend to think of knowledge workers and when you think of one, you think of one typically in a swivel chair and they always have a computer in front of them, so much of our knowledge work is kinda swivel chair processes over getting, you know, emails from customers and then we’re getting spreadsheets and then we’re doing some processes or paying invoices or processing claims and that person is going and swiveling around because we’ve got all of these proliferation of systems in our organizations.

That kind of work is boring, repetitive, and now we have RPA technology that can be doing that kind of boring work for them.

So I’m gonna give you a story of a typical example of how companies are choosing to deploy Robotic Process Automation… So here’s Sorry Charlie, he’s an HR specialist. And he is in charge of recruiting and on-boarding for his company. The recruiting part, he loves. He gets to go talk to business people. He develops job descriptions. He figures ‘Where am I going to market this?’ He likes to interview references. He likes to interview candidates. All of the parts of his job that require judgment, emotional intelligence, and human interaction.

The boring part of Charlie’s job is when he has to do the onboarding process. Do you know what it takes just to get an employee an office, security clearance, a parking pass, a Blackberry, a computer, gotta get entered into their systems so they get paid, all of that kind of swivel chair process gets done by the Robotic Process Automation, but I wanted to show you how Charlie has been transformed after automation. Isn’t Charlie happier? Charlie’s fitter. Charlie’s better looking after RPA…

There’s an important message in there about how companies that we study are choosing to deploy automation. They’re doing it to take the highly repetitive boring work out of a human’s job. It’s not replacing a human being, it’s replacing the boring, high volume, repetitive task, and liberating that human excess labor for other things.

Okay, so across the 14 of those studies that have deployed Robotic Process Automation, this is the value that was delivered to the organization and to some of the other stakeholders. I’ve been studying business practices for almost 3 decades with Leslie and I don’t know how many times I’ve ever been able to say that there’s a practice that results in a win win win, the win for shareholders, the win for customers, and the win for employees.

  • And I can say that about Robotic Process Automation because of the way the biased companies that we were studying chose to implement
  • if you wanna get this triple win I’m talking about, you’re gonna wanna follow some of the practices that our case companies get.
  • So the win for shareholders … we’re business people so let’s first talk about a return on investment.
  • Across our 14 companies, the lowest one year return on investment on their RPA implementation was 30% and that’s quite high if you’re a business person. The highest one we found was a one year return on investment of 350% one year return on investment.
  • So the rate of return on investment for the shareholder of the organization is very high…
  • Another example of a good business value is compliance.. In the companies that we study, compliance actually went up because you configure the software robots to follow all of the rules and regulations of your industry. And when you’re doing high volumes, it’s gonna do exactly how you configured it. When human beings are doing it, they get bored, they get tired, they might take shortcuts
  • customer value. So your external customers. Some of the services that get automated increased service quality and certainly speed of delivery.
  • So one of the companies in our study was Telefónica 02. They’re actually quite an early adopter of Robotic Process Automation. They adopted back in 2010. You might not remember this. It’s been so long ago.

But prior to RPA, if you bought a phone from 02, it would take a couple of days. After RPA, we’re talking less than 20 minutes. The other magical benefit to the customers is you were no longer calling the help centers saying ‘When are you gonna activate my phone?’ So they had 80% fewer calls to the help desk also unexpected after RPA.

So that’s an example of the value delivered to customers. you immediately think that people are gonna be afraid of automation

How they chose to do it, was they promised all of their employees ‘We are not gonna lay off anybody as a consequence of automation. We’re gonna redesign your work and liberate you and focus on more value added tasks that require emotional intelligence and human interactions.’

In the few instances of the companies that were looking to downsize, they did not lay off their internal employees. They either waited for natural attrition or in some cases, when we’re talking about hundreds of savings and FTE’s, they took it out of their service provider relationships ’cause that’s more politically acceptable than your own employees.

My favorite soundbite from our research is RPA takes the robot out of the human.

An example of the win win win The Associated Press. S

Journalists. News organizations. And you might be surprised to know that they’re a non-profit organization which I didn’t know and they’re based in New York City.

But non-profit does not mean you’re allowed to be losing millions and millions of dollars a year and that’s what was happening at The Associated Press. So they get a new CEO. and the agenda becomes we have to cover more news at a lower cost and protect our brand.

‘What do my journalists really like to do? My journalists like when I send them to Yankee stadium to cover a baseball game. My journalists like when I send them down to interview a chief executive officer at a large company.

What do they hate to do?’ Okay, what they hate to do is corporate earnings reports because the journalist would get handed that structured data.

So he decides that he’s going to automate corporate earnings reports. So when there was only the journalist, the human journalist doing this work, he only had enough excess staff to cover 300 companies each quarter.

And after automation, they now cover 4700 companies. So you see a huge, more than 10 times increase in the amount of news covered. You can now cover Canada, smaller firms.

  • London Insurance Market, decided to adopt Robotic Process Automation.
  • <They have to write a report for every client and customer> What you see up there, you submit the London Premium Advice Note. It has to be structured by someone like Amanda. She then finds errors, deals with exceptions.
  • And clearly the broker likes that to be done quite quickly because they like to see their money quite quickly.
  • So the Robotic Process Automation involved her being downloading really or many hundreds of decisions that she made and the many tax that she carried out configured into the software.
  • The software then was run with the data, structured data that was input into the process and she corrected when there were lots and lots of exceptions until eventually this piece of software could do this virtually 100% correctly.
  • So there is human involvement, dealing with exceptions, the RPA doesn’t structure the data. She has to structure it at the front end but the rest of it is totally automated.

The result that she used to process 500 LPA ends in 2 days, now they process 500 LPA end in 30 minutes.

That’s an incredible difference.

She anthropomorphised the software…. I said, ‘Why do you humanize the robot?’ She went ‘You gotta understand that it’s me. It’s partly me. You know, 400 things that I used to do are now done by the robot.’ And I see it has an enthusiastic trainee, a junior trainee.

This can scale.

So the earliest adopters in our study Telefonica 02 and the utility company, we did our interviews in 2015 had 300 robots running. That is 300 software licenses running and they’re running millions of transactions through this software per month so this can actually scale.

At the utility company, they have two human beings managing 300 software robots doing the work of 600 people.

And when you think about that. When 600 people were doing the work, do you need more than two supervisors? So the knock-on effects and the consequences for the future work you can see are coming.

Now these companies are also getting a tremendous number of savings, labor savings, and most of them are taking the labor away from their business process outsourcing provider.

Myth No.1, RPA is only used to replace humans with technology leading to layoffs.

Again, in our 14 RPA adoptions, none of that happened.

Business operations staff feel threatened by Robotic Process Automation.

We just saw the example of Amanda and that was very indicative of many of the employees that we interviewed during the course of our research.

RPA is only driven by cost savings.

We saw a much richer value proposition in higher compliance, higher service quality, faster services, employee job satisfaction increasing, so it’s a very rich value proposition

And finally RPA does not replace an entire person’s job. It only replaces the boring, broke, high repetitive, structure if-then-else activities in their job.

One interesting way of looking at the future is suggested by the McKenzie Global Institute. —- produced this graph which picks up an interesting point of key interest. It’s not jobs. That’s been a sort of mistaken focus really. But the percentage of the job or activity that is automatible.

We highlight examples of their study here which I know you can’t understand the graph. But basically what it’s saying is that if you take file clubs, 80-100% of the file clubs work is automatible in the near future. If you take a landscape and ground keeping work, 25% of that is automatible in the future. And even if you take a CEO’s work, potentially 20% or more of the CEO’s work is automatible in the near future.

More jobs will be generated

So there may well be opportunities for more work in terms of new data requiring, new data analysts, new services requiring new jobs, new combination of skills and jobs which can create jobs in that way.

And even the threats may be a source of new jobs. And one of the interesting aspect of that is the cyber security development. Let me just pick out that aspect… What is interesting about cyber security is that it’s both a solution to problems but also the technology created the problem in the first place and that might well be a future source of job gain.

We suspect an awful lot of jobs being created and anticipated and it feels like that.

And the social risks may well reduce the number of jobs but may well increase them in certain ways.

And the other thing to realize is that the course all that plays out in a social, political, economic, and environmental context. So it becomes quite difficult to predict.

There are many reports out courting figures on automations of jobs and they actually provide a very complex picture rather than a straightforward one job displacement.

  • For example, Deloitte 2016 sees automation taking over 114,000 jobs in the legal sector in the next 20 years. At the same time, it says the sector has actually grown 80000 jobs in the last few years.
  • While the Warrick University study says the legal sector will need 25000 extra workers between 2015 and 2020. New mix of skills are predicted.
  • McKenzie Global Institute is suggesting in the USA 60% of workers could have 30% or more of their jobs automated, 30% of US workers are in jobs where 50% or more of the work could be automated. On the other hand less than 5% of jobs could be fully automated.
  • A macro study by Forrester Research 2015 suggests in the USA automation will displace 22.7 million jobs, 16% of the total workforce by 2025. But also create 13.6 million, 9% new jobs. Meaning a net loss in the USA of 7% of jobs by 2025.

The first point I want to make is that we have studies we are seeing at the moment is merely the starting point. They are not to be accepted as definitive.

Point 2. There are many flawed assumptions running through these studies. They do not look at job creating. They assume the world of work and jobs stand still. They focus on jobs, not activities.

They have the worst case scenarios. They assume flawless robotic development. They do not allow for non-automatible human qualities. This is just a few that I would mention.

point 3, if robots are used on a mass scale for low level tasks and then for more cognitive, non-routine work, fewer people will be needed in those work categories. The FTE avoidance, redeployment and job loss will hit those work categories.

Fourthly, actually most of them are not quite getting what they’re really talking about. They don’t really define exactly what they’re talking about.

I think what is interesting is the total technologies that are going to have massive impact : social media and mobile technologies analytics big data and cloud Robotics Automation Knowledge Work which is what we’ve been studying, Digital Fabrication, and the Internet of Things.

Fifth point, everyone’s job, yours, mine, everyone’s will be transformed by at least 25% by automation by 2020.

Sixth point, during 2016 – 2025, our best guess … alright not to be stated as the LSE and University of Missouri professors stated that… Our best guess is that for every 20 jobs lost, approximately 13 will be created. We worked that through a little … at the organizational level projecting within organizations what was going to happen and that’s the sort of figure that seemed to get ellaborated.

Point 7, human qualities will remain vital in the future of work. Empathy, social interactions, specialist knowledge, experience, tacit knowing, leadership, imagination, creativity, composite skills, multi-tasking, teaming, all these things are valuable in the future.

  • We think that all the studies felt is taken into account three major additional sources of work creation. The first one of these is the Exponential Data Explosion that we are experiencing. On some figures, 90% of the world’s digital data which we tried to process was created in the last 2 years. how are we going to collect, process, analyze and use that data. We need more automation just to cope.
  • The second point, the other source of work growth is the Cross-sectoral Explosion of Audit Regulation and Bureaucracy. These themselves are amplified by the data explosion and the application of modern information and communication technologies.

We have the technology. We are busy using it and we really do have to use it with a lot more caution and thoughtfulness than we seem to be deploying mobile technology so far.

Technological change is not additive. It is decological, which means it changes everything. Our studies so far suggest that along with Michel Foucault on Technology, Knowledge, and Power, not everything is bad. But everything is potentially dangerous. That is why for the stakeholders in the technological future, each and everyone of us, there will always be work to do. Many thanks for your attention this evening.

And it’s also interesting the degree to which automation is full. We’re actually, I think I was reading a phrase the other day which I rather like. I thought which was ‘artificial artificial intelligence’. Either the notion that you got artificial intelligence. Selling is fully automated. But when you look at what is actually happening, you’ll discover that people are actually doing work that the automation can’t do although sensibly it’s been designed to replace people. So artificial artificial intelligence. I think you’re gonna see a lot more of that in the future because I’m not a great believer in the perfectibility of our technologies. There’s nothing in our history and I’ve been studying this with David Feeney and others and John Hindle for … since the early 1980’s when I was working in it and I’ve never seen really perfected technologies except on a very very small scale and the acceleration of the development of technologies, the desire to get it out there and get a return on investment doesn’t make for a perfect product.

Simon Bishop. E-Sale. Environmental Economics alumni. My question really is about looking at Thomas Piketty and his work on capital and labor, and he was seeing higher returns to capital increasing in equality. I’m interested in the distributional question, what implications does your findings have for the possibility that labor by becoming more skilled, if you like in contrast to the furthest idea of, you know, and de-skilling them thereby controlling labor through de-skilling. Are you likely to see more power coming to labor actually influencing in equality and just very quickly Haynes talked about a change in people’s working hours. He expected that we’ve been working 15 hours a week now when he struck late into the future productivity. Are we going to see a change in working cultures from a 5-day week to a 4-day week given that there … as you’ve said, we’re not gonna be replacing jobs quite as fast. Thank you.

It’s interesting that you asked that question having said Thomas Piketty because if you read his book, count the number of times he mentioned the words Automational Technology and you won’t get to five I suspect. It’s amazing analysis. Yes. You know, very very encyclopedic in many ways. But the factor of technology is not in the book. It’s just quite a staggering achievement. But we find that quite often with economists actually that they actually, you know, technology’s just another factor of production. And so it’s interesting that you have to ask that question. I mean I’ll back off on that answer because I haven’t got time but it’s pretty clear that all this will be through the filters of political, economic, and social factors. And that’s what dictates in equality or the levels of quality or whether technology is used to all command skills and amplify human strengths or the attempts to displace them. And that’s the best I can do for the moment.