Should automation projects be a business activity or an IT activity? What differentiates Bot 1.0 from 2.0 & 3.0? What is task harvesting? Why do less than 5% of organizations have an automation strategy, when automation is clearly such a major component of enterprise digital transformation?
As Chair of the IEEE Working Group on Standards in Intelligent Process Automation, Lee Coulter is in a unique position to answer these and many other questions about automation. Lee talks with us and shares his insider perspective of where automation is heading, what you’re really buying when you invest in an automation platform, and what the one essential skill is that will set apart successful individuals in a highly automated future.
Guy Nadivi: Welcome everyone! Our guest today on Intelligent Automation Radio is Lee Coulter, the CEO of Ascension Shared Services, and Chair of the IEEE Working Group on Standards in Intelligent Process Automation. Now, for those of you who may not be familiar, the IEEE stands for The Institute of Electrical and Electronics Engineers and the IEEE Working Group that Lee heads up is tasked with developing standards for, and I’m paraphrasing from their web page, the complex execution engines with fully developed management platforms, often coupled with increasingly sophisticated rules engines, analytics, machine learning, and cognitive computing, that are performing tasks previously requiring human operators. Collectively, this capability is known as Software Based Intelligent Process Automation or as abbreviated by IEEE, SBIPA.
Lee is also the Chief Intelligent Automation Officer of the Shared Services and Outsourcing Network, SSON. Recently, SSON published a report on Global Intelligent Automation with some very interesting findings. And so we’ve invited Lee to come onto our show and talk with us about the report.
Lee, welcome to Intelligent Automation Radio.
Lee Coulter: Fantastic, happy to be here.
Guy Nadivi: Your Global Intelligent Automation Market Report describes those embarking on process automation right now as “early majority adopters” which you distinguish from the “early adopters” who started three years ago. And this nomenclature sounds borrowed from the classic 1991 high tech marketing book titled, “Crossing the Chasm” by Geoffrey Moore. So if you believe we’re in the early majority phase of the market then does that mean you think automation has successfully crossed the chasm from the visionaries to the pragmatists?
Lee Coulter: Yeah, that’s a great question and I guess I’ll make a short answer and then add to it. And I would say, “Yes.” In the beginning if you look at the three unicorns of RPA, they’re all a decade or more old. So they’ve been out there for quite a while and the early adopters were 2010/2012 and in fact, we were one of the first, we were the first North American client for one of the big three RPA providers.
And at that time it was very much innovators and really learning how to do automation. It was a time when if you had a question there was really no one to ask. You got together and came up with the best possible path forward and we actually were involved in some co-IP development with some of the large consulting firms out there. And the chasm, have we completely crossed it? No. But are there lots of organizations that have? Yes.
Can I say that I feel firmly that we are standing on the leading edge of the early majority? I can. Two years ago a large consulting firm actually had published some articles about the fact that this was all a fad, and that it was going to go away, and 50% of them were failing and a lot has been done to understand why a lot of early implementations were not working out as well. And I think that there’s been enough experience over several thousand implementations worldwide that folks now know how to use this stuff. And get significant value from it.
Guy Nadivi: The report you put together is based on “real experience coming from thousands of successful, stalled, and doomed programs”. Can you please tell us about some of the more interesting automation programs you encountered that were successful?
Lee Coulter: Yeah, you know it’s always … I surely can. And what’s obviously more fun of course, Guy, is those that were doomed or explosively doomed in some cases. But those that were successful had a couple of key features that really made them different and it’s first important to understand that automation is a business activity. It’s not an IT activity. And I try to explain to folks that when purchasing an automation platform, you’re not actually purchasing a piece of software that has business logic in it.
If you go buy sales automation software or your ERP or your collection system or your payable system. These all come out of the box with business logic in them, and an automation platform is kind of like, it’s a box of potential and when that potential is guided by the business leaders, by the folks in the operation, a tremendous amount of configured business logic can now be put to work as digital labor in the enterprise. So successful programs first have business driving the program.
Second, I guess hallmark of successful programs is a mindful launch trajectory, and what do I mean by that? It means that the pilot was well conceived. A process that was chosen was not something that was regulatory or compliant influenced. It wasn’t a process that had 25 exception paths through it. It was something that was reasonably appropriate to prove that business could automate.
And the successful programs typically have a CoE (Center of Excellence) around them, which provides design authority, and change management, and interface with controls and audit, interface with IT. And all the supporting structures that need to make the program on an overall basis successful. So those are some of the successful programs. This is not … It’s not a panacea. It’s not a get-rich-quick scheme. And it’s not, as some would have us believe, something that you just open the box and automatically you’re automating stuff.
This is a serious technology that can perform serious duty in the enterprise but it does require competent management of it as a program to be successful with it.
Guy Nadivi: You alluded to some of the doomed ones and I can imagine you’ve probably come across some real humdingers and in fact, in your report you explain that the greatest cause of failures in proofs of concept or pilots, can generally be traced back to one of two major mistakes and both have the same root cause. Can you talk about those a bit and how they should be avoided?
Lee Coulter: Sure. In the study of 88 failed programs, there were really two populations of failures and in one of them in about half of that population the failure was directly because IT led the program. And that’s not to say that they were not smart people, that they weren’t a good IT function, but they weren’t the business. And automation is about digital labor. It is about executing the operating processes of the enterprise. And that is not IT’s role.
And so when IT approaches a software purchase, they develop an RFI, and they do a beauty pageant. Then they do an RFP, and then they compete everyone, and then IT procurement gets involved, and it really becomes a very expensive, very public labor-intensive effort. And then because an IT project has to have an ROI, well the most complex process was chosen that’s 160 steps long with 20 different exception paths. And so you go through six or nine months of procurement, and then another six months of building, and then it doesn’t work, not surprisingly, because a whole lot of “becauses” are in there.
And the second major population, which was about 80% of the other 50%, it was because the business treated the program like a standard software project. And here again, same series. And now in this case we had appropriate leadership and sponsorship from the business, but with a lack of recognition that this is a business activity, that the creation of configured pieces of automation and the management of them is very much a business activity and it requires your subject matter experts to be significantly available to the effort.
It requires, if you’re an agile shop, it requires regular access to the users and the processors, and the analysts, and agents of the organization doing the work. So really, as I mentioned a little bit earlier, the fact that this is a piece of enterprise software that doesn’t have any business logic per se in it is really fundamentally different, and it requires a different programmatic approach that really … this is the sort of thing the chasm of despair really was filled with, a lot of these failures that can be traced back to these kinds of same fundamental faults.
Guy Nadivi: Lee, you’ve described the evolution of intelligent automation’s history as a 3-tiered progression from Bot 1.0 to Bot 2.0 to Bot 3.0. And you say that Bot 1.0 is the early years which includes focusing on automating basic, repetitive, recurring activities which you refer to with a great phrase, “task harvesting”. Bot 2.0 is a more mature state with additional tools, increasing numbers of bots, and some experimentation with cognitive solutions. Can you please describe what you believe is Bot 3.0, where we are today, and how it will benefit organizations deploying automation?
Lee Coulter: Sure. I do want to emphasize that there’s nothing wrong, and in fact I really do encourage folks to move through 1.0 and 2.0 to get to 3.0. Jumping straight to 3.0 is a risky place to be.
There are a lot of organizational capabilities, and organizational learning as you move through the early efforts of task harvesting. Here again, task harvesting is wonderful. Saving money and learning how to automate is important, also as you build your CoE.
When you start to get into Bot 2.0, now you realize that you don’t have as many straight-through processing processes as you thought. As you begin to do the really detailed kinds of business analysis for automation, which it’s a specific discipline. It’s a new discipline. If you follow the CMM process levels, it’s a Level 7 process discovery, which is very, very different. Part of the enterprise learning here is exactly how many times a process starts and stops, and how many different operands and condition sets are needed in order to get a process on an end-to-end basis, through to completion.
You start in Bot 1.0 with picking just chunks of a process that are straight-through processing, or STP. Then when you get to Bot 2.0, you’re really looking at, OK how do I start to engage a rules engine? Can I pause the automation and execute a set of queries, do some assessment against what was returned from the query in order to determine a current condition set? This would be your basic decision making capability in automation.
Bot 3.0 is just really beginning to, even for the early adopters, this is fairly new stuff. It basically, Bot 3.0, is when you have sufficient data to build predictive models that reach a level of confidence and a sufficiently acceptable level of false negatives and false positives that you can use that prediction on the next best action to actually take the next best action. Here again, it’s an iterative process. This isn’t a case of where you build one model and everything is now being aided by machine learning. This is a case of where you’re really, through deliberate planning and execution, begin to introduce prediction and prescription such that over time, some of your most basic decisions for next best action can now be taken over by technology.
I do believe that there will be a Bot 4.0 as well in which certain knowledge domains will be … The data models will be licensable, as opposed to needing for folks to build them. But once we now cross the line into Bot 3.0, what we really have now is the ability for predictive models to begin to orchestrate automated tasks without human intervention. That’s really, that’s the holy grail. That is the intelligence process automation, as opposed to task automation.
Guy Nadivi: Yeah, I think that automation is such a different paradigm, it really requires a different way of thinking, and that really leads into the next question I have for you, which is regarding the fact that your report stated very few organizations today have a Head of Automation. Without one, it will be difficult elevating automation to a strategy from a tactic. In fact, you estimate that today less than 5% of organizations have an automation strategy. What will it take to change market sentiment so that we start seeing more Chief Automation Officers in the enterprise?
Lee Coulter: This is a really interesting one and was kind of a surprise. But once you learn it, then you say, “oh well that kind of makes sense.”
I’ve had a long standing saying as a matter of leadership, which is if you want something done, make it someone’s job. This is very much a case of that idea. In so many enterprises, automation got a foothold in the operation probably somewhere in lower middle management as a tactical response to the relentless incremental cost reduction expectations. You need to take out that … You get no inflation and you need to give me 5% year-on-year, and how are you gonna do that?
In a lot of organizations, automation was brought in as a way to harvest those tasks, introduce digital labor to the enterprise, really as a way to satisfy that need for consistent year over year productivity. What happens to organizations as they begin to mature, they realize that digital labor is capable of a whole lot more. There’s an interesting inversion that happens here.
You start off by asking yourself, where could I use human labor, or digital labor, to do the things that human labor is doing today? When you invert that, you ask the question, where could I deploy digital labor to do jobs that no human could or would ever do? That opens up a box into the world of strategic automation where by use of automation you may change customer experience, which may change the costs of your customer acquisition, may reduce the customer turnover, may improve customer experience at the most basic level, can impact your reputation.
I use the example when a single defect can have a reputational impact, and we’ll use Uber as the example or Tesla, where single defects have a very significant impact. Automation can be used to deliver those sorts of strategic benefits that typically happens after you have achieved a certain level of financial return with it. But what we see is that there’s a correlation between the organization installing a person whose job it is to put automation to good work, to deploy digital labor. When it’s that person’s job, they’re exploring it from both the tactical and the strategic perspective. They’re considering speed, quality, experience, obviously cost, but asking the question about what is the impact to efficiency, and what is the impact to effectiveness of a process? What could digital labor do? That’s really what we see is if you have a Head of Automation somewhere, you’re more likely to have a strategy. Same thing goes if you have a strategy, very often it is a person who has been tasked with leading that strategy.
Guy Nadivi: I found one of the more interesting perspectives your report offered was with regards to procurement departments. You advocate procurement departments start seeing themselves as creators of business value and profit drivers themselves, rather than just as a support function. You advise them to stop viewing automation as an enabler for a specific project, but rather as a catalyst for enterprise-wide digital transformation. In your experience Lee, how prepared are typical corporate procurement departments to adopt this shift in mindset, and can they be incentivized to do so?
Lee Coulter: You know, Guy, this is a really … It ties back to the earlier dialogue with the core reasons that a lot of programs fail. I literally have seen examples where organizations have spent 400 to 500 thousand dollars procuring a 200 thousand dollar piece of software. And because I described an automation platform as a box full of potential, procurement organizations really don’t know how to buy a box full of potential. It’s kind of like buying a box of random Legos, and your kid promises that they’re gonna build an Eiffel Tower, and a model of the White House, and the Golden Gate Bridge, and it’s gonna generate all this value. But really, it’s a tool in the hands of the business.
Procurement organizations wrestle with how do I know what’s an appropriate price? How do I compare licensing models? How do I compare actual technology capability? Interestingly in the new IEEE standard 2755.1, which we are working very, very hard to get published before the end of this calendar year, even to be able to say what differentiates one product from another is a bit of a challenge, and the new standard will bring a lot of clarity to that, defining about 165 features and functions of this category of tool in general.
Coming back to your question on procurement organizations, this is another place where the business really has to drive the process. If the enterprise lets the procurement engine, it’s like putting a train on a track. You know where the track leads, but in this case, you need the track to go somewhere else and you might have to lay some track. So, it can, procurement organizations can rise to this challenge, but they need to do so consciously and with the close alignment of their business sponsor. If the organization already has an IT procurement specialty organization, as many large enterprises do, looking for the right procurement leaders to educate as to the difference in this exercise before the process begins is critical.
Guy Nadivi: There’s a section in the report also about automation anxiety, or something I started referring to with the term robophobia. And, you advise that the way to win over resistance within people is by focusing on influencing rather than conquering them. Lee, can you please describe how best to win over someone who sees automation as an adversary rather than an ally?
Lee Coulter: So, this is…it really, Guy, goes back to the fundamentals of organizational change management, or just human behavior in change management. And what we’re doing here is we’re changing paradigms for people. And, in the beginning there’s lots of excitement. It’s like oh that copy and paste thing that I used to have to do I won’t have to do that anymore, and I’m so excited that I won’t have to do that anymore. And then as time goes by people begin to realize that as people are leaving that department isn’t back-filling anymore. And, every two to four weeks I’m now having to be trained on how the new process works because incremental portions of it have been automated. And now I’ve got a supervisor who is responsible for the output of a process or group of processes who used to be able to know how work was going by going and talking to her team, and being able to see the work being done. That same supervisor also had a sense of value and a sense of contribution based on the number of people being managed.
So, we’re fundamentally changing the nature of work in a way that has a result in something called “cognitive load concentration”. But people that had very transactional jobs suddenly find themselves confronted with the likelihood that they’re going to be asked to do more complex work that has a greater cognitive load. We’re finding that we’re asking supervisors and managers to be responsible for both the humans that they manage that they can see, and the digital labor that they are managing that they can’t see. And, this whole thing is automation anxiety, and it’s very real. And, there are even some charts and graphs out there that kind of show you over time when you should expect to see it. And, it is literally a limiting factor in how quickly automation can be injected into an organization.
So, when you begin a program having your messaging already thought through. Guy, I know hundreds of people in this industry, and outside of some commercial BPO’s I don’t know anybody who’s running a program that has resulted in job eliminations. And, being able to say that to your folks as you begin is a key part of messaging. Talking about retraining, and upskilling, and how the organization will keep your folks relevant as the works changes, these are really important parts of change management and getting the human behavior into this new world of rapidly changing delivery process and this mix of human and digital labor.
Guy Nadivi: And that’s a great lead-in to the last question I wanted to ask you about reskilling or upskilling. You advocate for organizations to include reskilling or upskilling in their automation strategy, as it will instill confidence in the people being asked to incorporate automation into their work. What about the people who want to change their roles by actually delivering automation to an organization? What do you think are the most important new skills they can acquire to make this transition?
Lee Coulter: So, this is a fantastic question because it really speaks to the changing face and work mix that we’re now facing. If you look at the literature from 2014/15, it was 169 million people are going to be out of a job, and here we’ve ranked 700 different jobs in terms of their propensity to be automated away. And that rhetoric has pretty much gone away and been replaced by a much more sane and rational analysis of the tasks within a job that are automated, or the tasks within a role that are automated, or the roles that a person may play throughout the day which have some level of automation potential.
So, no question that becoming more familiar with automation and the role of technology is important, but I would recommend to people that they become versed in process improvement capabilities. Whether it’s Lean, or Six Sigma, or Kaizen, or it’s a business analysis for automation, they all have in them a lot of fundamental tools and skills that cross over here. But, the successful contributors of the future are going to be able to understand how a process gets done and understand how different kinds of optimization can be put to good use. Whether that’s automation, or information, analytics, decision-making visibility, etc. So, becoming more versed in process management overall, learning the skills of the next job up on your job ladder. Whatever that is. Beginning to pay more attention to that. In some cases that means folks will have to take a more active hand in managing their own skill base if their organization isn’t already planning for reskilling and upskilling.
There also needs to be an organizational conversation about the willingness to make an investment in your people. So, some parts of an enterprise are very significantly automatable and you will need significantly less folks in those functions in the future. So, is the organization willing to have somebody work part-time while they return to school to get a certificate, to learn a new skill? Will the organization invest in trainers to come in and on company time give associates and employees new skills? So, I guess becoming involved in the work of process management and process improvement, or looking up the value chain in your own career ladder in your job family are the areas that I would urge folks to concentrate on new skills development.
Guy Nadivi: Alright. And with that it looks like we’ve run out of time for this episode of Intelligent Automation Radio. Lee, thank you so much for joining us today and sharing your thoughts on the global intelligent automation market. It’s been very interesting and I’ve really enjoyed having you as our guest.
Lee Coulter: Thanks so much Guy. I really had a good time.
Guy Nadivi: Lee Coulter, CEO of Ascension Shared Services, Chair of the IEEE Working Group on Standards in Intelligent Process Automation, and Chief Intelligent Automation Officer of the Shared Services and Outsourcing Network. Clearly a very busy guy. Thank you for listening everyone, and remember – don’t hesitate, automate.