Work Minus Robophobia with Neil Ward-Dutton

13 Aug 2018   |   Technology

Work Minus Robophobia with Neil Ward-Dutton

13 Aug 2018   |   Technology

Welcome back to Work Minus, where we talk about what we need to drop from how we work and quick pivots you can make today to get closer to the future of work. Today, our guest is Neil Ward-Dutton. He is the founder at MWD Advisors and this episode is Work Minus Robophobia. Hi Neil. How are you today?

I’m good. Thanks, Neil. Thanks for having me on.

Yeah, it’s always great to have someone of your caliber on the show. You’ve done a lot of writing and researching lately, so we’re excited to share your insights. Why don’t you start with telling us a little bit about who you are, and what types of things do you do?

Okay, cool. So, I’m Neil Ward-Dutton. I run a boutique industry analyst firm focused on how digital technology changes work. And fundamentally, that means figuring out how new tech is really changing the way that people make decisions, the way that work gets done and coordinated, the way that knowledge gets shared and the way that customers get the outcomes they want. And a big part of that is all around automation and AI. And that’s what we’re gonna be talking about on this call. I’ve been doing this at MWD Advisors for about 12 years. And for a number of years with other firms before. So, I’ve been around the block a few times as they say.

Yeah. And I’m sure you’ve seen a ton of changes, especially in the recent years. We talk about AI on the show a lot, but one of your specialties is also RPA or Robotic Process Automation. So why don’t you bring us up to date on what RPA is?

Sure. Yeah, so Robotic Process Automation or RPA is probably the most aggressively promoted and explored topic I’ve come across in a long time. It just seems to have gone over the last two years from very much a kind of niche concern to pretty much everybody willing to talk about it. To try and give you a very quick stir on what this is, at its most basic level, it’s pretty simple. It’s fundamentally a software that mimics the actions of a human, against, some kind of legacy system or some kind of system that’s not really amenable to more traditional kind of programmatic API based integration. So RPA systems typically will have a design interface where you capture and model the actions of people working against the screens of legacy systems. So, calling up a screen, typing in some data, running a query, getting some results back, running a report and so on.  

And what you can do with these systems is you can aggregate those, right? So you can work not only against one system, but you can actually have these programs or robots as they’re called. You can have them maybe interacting with one system, maybe getting data from a spreadsheet or from a database, maybe sending an email, or running some kind of automated decision in the background. So actually there are some parallels with what we’ve seen with workflow. A design environment where you can specify actions and flow between actions and decisions. But specifically focused normally on very, very particular kinds of tasks, and those tasks tend to be, as you might imagine, very routine, often kind of clerical tasks, which has to do with simple kind of querying and updating of backend systems. You do find a lot of this kind of work in large organizations, still a lot of this done by a human clerical workers where it’s very routine, very rules-based and very repetitive.

So, is RPA mostly used in an enterprise environment where there’s a lot of people working on these types of things?

That’s right. Actually, Neil, it started out in the outsourcing environment. So you saw big outsourcing companies who were already aggregating work from multiple clients, and really looking for ways to improve their efficiency and improve their profitability. So they were already, in a captured way, managing a lot of very routine work on behalf of their clients. And RPA technology really started out as a way to take some of the most robotic work out of the hands of those people running the outsourcing contracts. But it’s spread from there into back office and shared services environments in banking and retail and utilities and pretty much everywhere. And it’s also starting to spread into places like call centers as well.

And so now we’ve mentioned AI before, how has AI influenced RPA?

Yeah, that’s a great question. Really the primary way in which AI related tech has started to be used in conjunction with RPA is kind of capturing some of the upstream work that typically needs to get done in an RPA context. I need to be more specific about that. So a kind of stereotypical scenario that you would have looked to address with RPA is, dealing with very straightforward inquiries or posting of information. So a great example is around accounts payable, where if you think of a really large telco or utility or a manufacturer or something, and you think about all the invoices they might receive every month. There’s going to be thousands upon thousands upon thousands of these. And there needs to be a process for receiving those, figuring out whether they’re correct, whether all the data is complete, reconciling those against other systems of record and then paying those invoices and communicating back to the supplier, right? So, that’s a very, very common use case for RPA. And organizations start out by getting the invoice data, and then doing the simple process of kind of matching the invoice against the purchase order and checking it against one or two other reference systems. And then making the decision, paying the invoice and so on, all right. So that’s the kind of backend piece.

But where AI can be used is taking some of the frontend piece of that as well, because what you very often find is even in really large, really well established companies where they might already have big investments in kind of B2B electronic kind of networks for full financial processing, a significant proportion of invoices they receive are still going to be sent over email, or they might be a PDF attachments or even it might be via fax. And in that situation, like you have, think of it almost like a multi-channel problem. You’ve got lots of documents coming in, in different formats, every day different invoices, some of them maybe things sent in error, either accidentally or maliciously. So what you find is that more and more organizations are using intelligent OCR capture technology to take incoming documents, let’s say PDFs or scans something from a digital mailroom, and then using machine learning models to pattern that and say, okay, this is an invoice of type X or type Y. I know what that looks like. I know, because based on past experience, that the line items are kind of at this point in the invoice, and the total is over here towards the bottom right, and the PO number is in the top right, not the top left because it’s from supplier x.

And you’re able to use intelligent capture, AI tools to kind of categorize, to extract data, to clean the documents up, to do fuzzy matching and solve problems if it’s not entirely clear where the invoice has come from, for example. Sometimes even we see organizations looking to do machine translation on documents. So, actually maybe if documents perhaps not invoices but other kinds of documents come in and multiple languages, translating those all into English, for example. So, slightly long-winded response, but I guess what I’m trying to say is these technologies get used often to kind of provide sentencing and categorization capabilities onto the front of robots, if you like, that makes sense.

Yeah, yeah, absolutely. Obviously the benefits are pretty straightforward. You’re talking about speed, you’re talking about accuracy, you’re talking about reducing your errors. Just taking a lot of this work and giving it over to, in some sense, not a robot, like the type you would see that’s built by something but a robot that’s working behind the scenes behind the screen, so to speak. So what are some of the other important benefits that people may not think about when they think about this RPA and AI working together?

Well, the key, you mentioned, many of the main ones, right? And there’s kind of like a higher level benefit that you can get from these technologies. High level on top of what you’ve identified any all-around speed and accuracy and errors, which is their own customer satisfaction. So, when you are automating aspects of kind of customer facing processes, and you’re able to deliver greater speed, reduced errors, more accurate results, clearly that’s gonna translate into a better customer experience and a higher customer satisfaction.

And there’s a number of cases you can find where organization is about great results there by just being able to respond to customers so much more quickly and more transparently. Building all that transparency thing though something else, which very often as part of the calculation when people think about ROI and business case is regulatory compliance.  Because in so many cases, you find, particularly in back office situations, the stuff that we don’t necessarily see or you can think about in large businesses. A lot of that clerical administrative work, a lot of it’s kind of opaque. It’s difficult to really see what’s going on. And a lot of the compliance stuff that goes on is actually armies of people with clipboards chasing around departments saying, okay, did you do that? Did you do that? When did you do that? Did the right person do x or y? When you start to automate some of these slightly more opaque parts of a business operation, you get a huge amount of transparency. So it’s easy to see, not three months later. It’s easy to see right now what work is being done? Where? What kinds of rules are being followed? Is the process that we identified is that big followed? Well, you know what, if it’s been automated through robotics and it’s probably following the process we designed, right? Because these things just follow rules. So that’s a huge deal. Particularly in financial services, but actually in many, many other situations as well. And you think about things like, I’m under the GDPR, the EU general data protection regulation. Think about the rights given to people to make requests of companies. Tell me the information you have on me or remove all my information. Those are great applications potentially for similar kinds of automation.

Yeah, absolutely. We’ve titled this show Work Minus Robophobia, so we don’t want people to be afraid of robots, as they are. Why do you think the future of work is humans and robots, not humans or robots?  

I’m really glad you asked this question because I see so much confusion about this, and it’s not helped by a lot of headlines which are about, robots are going to take our jobs. Another headlines that say, robots aren’t going to take our jobs. And to me, they both count as a kind of shouting across the fence at each of the are actually taking aim at the wrong thing. Because from all the work we’ve done, what we’re finding is that the impact of robots, whether you’re talking about blue collar or white collar. It’s very, very rare actually, that you’re already looking at a whole job where something is gonna be automated. The impact is principally actually at the level of individual tasks. And there’s very few jobs which are only made up of one kind of task.

So, there are gonna be across a great number of jobs. There’s gonna be some tasks which are increasingly amenable to automation, whether that’s RPA or whether it’s an API driven integration or workflow or some use of AI as part of all that stuff. Some tasks will be increasingly amenable as the technology matures to more automation, but it’s gonna be an awfully long time before all tasks that most people do are amenable to automation. So in the vast majority of cases when you think about jobs and roles being done, we’re gonna see people and software collaborating. And there are many, many tasks that require real expert discretion, problem solving, creativity, or indeed that may well be ethical or regulatory reasons why we just were not comfortable with having software make certain kinds of decisions. I think, if you look at the factory environment, for example, where yeah in some factories there has been an awful lot of automation.

But then if you take a step back and kind of blur your eyes a bit and look at what those factories look like, the reason why there is so much automation in them, is because actually what’s been done is the environment has been very strictly controlled. So the story of automation in that context is not just about the robot. It is actually about the environment. The environment has been designed as much as the robot’s been designed. So you think about how these robots move through the space. And very often they’re following very particular paths that have to be clear. If they’re not clear, everything goes to tailing in a handcart or robot swinging around in space, physical space. People have to be kept in very particular kinds of environmental area, otherwise there could be collisions, right? So, it’s just a reminder that actually if you’re really going for high levels of automation, it’s as much about designing an environment in a very constrained way, as it is about designing the automation. And so that does place restrictions on what you can realistically automate today and maybe in the next 5, 10 years. Because the real world can’t be designed like factories. We can’t control the context and the environment for work in that way, apart from, in some very specific contexts.

Yeah, I know, absolutely. It’s fascinating to think about what tasks instead of jobs are gonna be taken over by certain types of robots that come through. Do you think the trends can be there for a long time? Do you feel like there’s gonna be a golden age when humans and robots can work together on these things, but then eventually we’ll they’ll figure out a way to code all these things or we think are uncodable. What, – looking out, how far in the future do you think we’re seeing here?

Well I think, for the next 10 years it’s dangerous to think about these things until they go on record. I think the next 10 years is gonna be continued progress, gradual maturing of the technology to address more and more kinds of tasks. But I guess where I’m more cautious is where we get beyond that, 7 to 10 years because actually the technology is changing so rapidly. It’s very, very difficult to say with any kind of certainty what kind of environment we’ll end up in. We could end up in an environment where most work is automated. But you know what, if we were to be in that environment, what would we do about unemployment? What would we do about salaries? What would we do about our economy? How would that work? So I think there’s lots of opportunity for things that aren’t anything to do with technology to get in the way. So for example, labor unions or politicians, or international competition, all kinds of things that aren’t really to do with tech but it also can produce something to economics, quite possibly going to kind of derail or shift the agenda around automation, it’s pretty uncertain in the long-term.

No, absolutely. So let’s bring it back instead of 10 years in the future that’s bringing back to right now, let’s talk to somebody who’s running a procurement team or was on the marketing team, who’s a manager in a business right now that wants to take the first step towards something like either RPA or some version of that. Where’s a good place for them to start?

It’s really, really important. In fact, I was talking to somebody just this morning, about this is really, really important to understand that RPA is one specific overall piece of the puzzle around automation. It’s particularly useful when you’ve already been through some kind of probably some process re-engineering, and you’ve already identified work that is done in many places, but it can be standardized and it can be centralized. So what you really want to be looking for is places in your organization where there are specific teams who are already doing pretty significant amounts of the same kind of work. And that work is probably very routine, it’s very rules based, it’s very prescribed. So there’s probably some kind of manual sitting somewhere saying, this is how you do this work. You go to the system, A, you type this information, you get information back, you go to system B, you look up a certain reference code, that kind of procedural level of prescription.

If you can find departments or teams where you’ve got significant work of that kind being done, that is a great place to start. If you try and apply RPA to a situation where you have one person doing little bits of work here and there, it’s really not so well defined. It’s gonna be quite hard and take quite a long time for you to really get a return on your investment because actually, RPA does need the work to be designed in very particular ways for you to get the right kind of return. So that’s really the best place to start is find places where the work has already been kind of re-engineered and streamlined. And where it’s already pretty predictable and prescribed. That’s where you should really get started.

Well, excellent. Neil, we can talk about this for several hours I’m sure, but this has been a great start to get somebody to think about RPA. Think about how AI is influencing that and everything. Thanks so much for being on the show. Where can people go to connect with you more?

Oh, well, it’s been a pleasure. Thank you so much for the time. Best place for people to get in touch with me and find me on Twitter really, and my handle is easy. It’s @neilwd.

Yeah. It can’t get any easier than that, right?


Well good. Thanks again for being on the show. We appreciate it and we hope that everyone enjoys this. We look forward to connecting with you more in the future.

It’s a pleasure.


Subscribe to The Digital Workplace

Join the journey to a better future of work