Director, Research and Thought Leadership
KPMG Global Management Consulting
Shared Services and Outsourcing Advisory
STAN: Dave, obviously, robotic process automation (RPA) is a big topic in the market today, whether from the perspective of process automation versus outsourcing or from the broader standpoint of the impact it will have on white-collar labor. Cutting through the hype around this topic, give us your perspective on the reality of process automation today and, for a client organization, how high should it be on its priority list?
DAVE: One of the questions clients increasingly ask me is, “Is this hype or reality?” so it’s important to differentiate between hype and reality and what is new versus old because many forms of process automation are not new and are clearly not hype. The basic building blocks of many RPA tools, however, are capabilities such as screen scraping, workflow, and rules engines that have been in the market for many years. There are plenty of companies that have successfully implemented various forms of RPA. There are mature solutions in the market and documented efficiencies gained and savings achieved. You have the [IBM] Watsons, you have Blue Prism, IPsoft, Automation Anywhere, and many vendors and providers that have built businesses around automation, so there are plenty of solutions that demonstrate it’s not hype.
Hype comes in with the new, more advanced forms of RPA based on newer technologies and capabilities such as cognitive computing. RPA is built on technologies that have been in the market for many years but more recently there have been significant advances in very key areas such as NLP (natural language processing), significant growth in basic computing power, and major advances in capabilities to process unstructured data that are enabling RPA to rapidly evolve. This last point is critical because over 90 percent of the world’s data is unstructured. The capability to rapidly and accurately make sense out of this data is a huge advancement. There have also been great advances in machine learning, that is, contextual learning in which an environment or system is trained to understand a topic and then can independently get smarter going forward.
Relative to business process and IT outsourcing, RPA is and will continue to have a great impact. Today, clients are increasingly asking, “Should we automate or should we outsource?” Automation is going to happen either way. It’s just a question of whether the service provider enables automation or a client does it internally or both. When considering utilizing a provider, key issues to address include how and how much is it adopting RPA, what benefits can it bring to its clients, and what are future plans for expansion of its usage. The alternative for a client considering expanding deployment of RPA internally involves asking does it have the skills and appetite for change, does it want to take on vetting and implementing technologies, what are change management issues to address, and how does it organize to support these efforts—via the IT group, business units, through an RPA center of excellence, etc.
STAN: Continuing the discussion on the impact that RPA is going to have on the outsourcing industry, will its growth and maturation decimate some firms that provide services easily automated? These firms will have a lot less staff, but will they have higher profitability and can they manage the transition from labor to digital? How much of this scenario is a near-term reality that’s impacting service providers already, or is the scenario that enough of the work they perform is still customized enough that RPA will not have a huge near- to mid-term impact? And a follow-on question from the client perspective, should they really care about the business model of the service provider and RPA’s potentially negative impact on it? Or should buyers have concerns that RPA could threaten their providers’ economic viability and therefore their ability to provide the services clients still need?
DAVE: When it comes to the service provider, it is a simple case of automating to survive. Providers that do not aggressively embrace RPA will soon be unable to compete with their brethren that figured out how to exploit RPA across their operations. This is especially the case with labor-based offshore outsourcing where we are already seeing diminishing returns due to rising wage levels and dwindling labor arbitrage benefits.
Automation can bring providers tremendous benefits such as instant scalability, improved activity and accuracy, greater process consistency and predictability, and hedge against rising wages and high turnover rates. There is also enhanced auditability such that a provider has the capability when a problem occurs to examine logs and determine exactly what happened. This is not to say that every activity is fully automated, but where applicable automation offers providers great benefits there are many practical challenges in adopting and expanding efforts and this is where many providers will stumble and some will fall.
Many providers are already actively utilizing more advanced RPA services, though often still in pilot or experimentation efforts. We are working with providers proposing or utilizing more advanced forms of RPA to our clients and having discussions around, “What does my client get out of that? How do they benefit? What do they see that different and better?” If RPA is behind the scenes and through its usage providers get greater margins, how do clients benefit from these gains? Service providers must automate but the real question is, what do they put forward to your client as a result of automation, what is in it for them, and at what cost?
STAN: If I’m a client considering the adoption or expansion of the use of process automation, what are some of the key things I need to consider? If I’m a client assessing RPA opportunities, how much should I focus today on established technologies and services and how much on the cool emerging areas of RPA such as IBM’s Watson or IPsoft’s Amelia?
DAVE: You’re talking to an engineer, so “cool” is always fun but not necessarily beneficial. When you look at the opportunities for automation, in some ways it’s similar to assessing outsourcing opportunities. So to the questions, “What should I be considering? What could I automate?” I answer, “Is it a task repeated frequently? Is it fairly predictable? Is it a process that’s based on fairly well-established rules?” That’s the kind of thing that you see in outsourcing as well, along with, “Does the work vary in volumes? Does it scale up and down?”
RPA also adds some additional dimensions, such as is there an electronic trigger for automation. By this I mean you need something that’s going to start an automaton, so you need to have the ability to trigger it based on something that starts the process. Another point to address is are there multiple technology systems involved and are they adequately integrated and interconnected.
But to your point, there is a very broad spectrum of automation tools and platforms and capabilities. On the base level of RPA capabilities, for example with screen scraping and workflow, clients can deploy these tools and gain benefits fairly quickly, though there is the requirement to build supporting workflow manually. It is also not necessary to automate an entire process; good benefits accrue from partial automation. Some tools, especially those supporting IT services and that have been in the market for several years, come with often extensive configuration and functional knowledge out of the box.
There is definitely a requirement for larger investments when it comes to higher end RPA services such as cognitive computing. This investment is in dollars for the platforms and the time to develop them. Even the best of the cognitive tools have contextual learning needs in that it is necessary to “teach” the system in a defined topical environment first before it has the capability to get smarter and learn independently.
Look at what IBM’s Watson does, for example, in the medical and the oncology world. Watson did not come out of the box understanding cancer. First, it was necessary to teach Watson about cancer, which required strong domain knowledge (i.e., what are the trusted sources to use for teaching) as well as the legwork to teach (i.e., where to get content, how to filter, what are the privacy issues, what’s in it for the contributor). It was then necessary to teach it prognosis versus diagnosis and understand, for example, what are the key characteristics of a tumor. Once this was done, it was possible to set Watson free to analyze millions of images and millions of case studies and medical records. But it’s that contextual learning that can take years of perfecting before you set it free and see the benefits.
So the benefits you get are going to vary greatly with which way you enter. Most of our clients will or already have adopted RPA via base level capabilities such as screen scraping, workflow, and process automation type of tools that provide knowledge and value out of the box.
STAN: So for most clients it is “walk before you run” and have realistic expectations about how long it’s going to take to be able to achieve some of the bigger benefits of this technology. But are there other risks organizations should be aware of when they are considering adoption or expansion of process automation?
DAVE: There clearly are risks clients need to understand and address proactively. RPA is not a panacea. It represents some phenomenal advances in computing capabilities and has tremendous applicability, but it’s not a panacea for all problems nor will it always work as planned. One major risk clients face is that RPA is not a “drop in and walk away” effort. Just as you cannot, or should not, hire employees and leave them alone for the next 50 years and assume they’re going to do their job as required. Just as every human employee requires ongoing training and support, so do automaton efforts. The automaton has initial knowledge that its owner must maintain and enhance over time. This is why governance is important as a discipline and group. Once automation in enabled, there are requirements to maintain required enhancements, make changes to logic to reflect changes in the business, implement new technical capabilities, and respond to lessons learned. These are all elements that a governance group must ensure are addressed. Automata do what they are told, and if what they are told is outdated and inaccurate, the automata will start to do the wrong thing until corrected. Performing the governance role to address this ongoing need is one of the key roles for humans to still perform as more activities are automated.
STAN: So to wrap up and summarize, we still will need to keep a few people around going forward to do training, governance, and perform other activities you just described and to interpret and validate the results of the output from automated systems and processes. But Dave, what is your perspective on the long-term implications of this from an employment standpoint? Should this be a worry for the white-collar workers of the world, particularly those in jobs such as audit, or is the scenario that in the future there will remain a strong need and demand for humans, especially to really help to craft automation technologies and systems into something that can provide meaningful business benefits?
DAVE: In some ways this is a little different than the industrial revolution that entailed getting rid of some of the manual labor and moving workers up the food chain in terms of activities performed. One thing different with automation is that it attacks all sections of the food chain. There is definitely a major potential employment impact across a broad range of occupations. It is not just a situation of “I have 100 people doing this. Now I’ll have these 50 go off and be fully automated and I’ll repurpose them to run the automata.” This job shifting is not viable for some to many workers, and as automation systems advance less human intervention is needed. Ultimately, the ratio of automata to humans will be anywhere from 5:1 to 15:1 depending on the activity.
Employees are needed to maintain topical knowledge, vet sources on input for quality control, and oversee the automata. Just not a lot of them! There is the opportunity to free up members of the workforce, similar to the case when organizations outsource, to do more strategic work, but there is going to be a drought in job opportunities when the automata are fully implemented across all aspects of the hierarchy of responsibility.
It’s not a gloom-and-doom scenario. It is not my impression that people are going to be faced with tremendous unemployment, but I will say it is not as simple as trimming the bottom and moving people to the top, because the jobs at the top also are going to be impacted by automata which are doing work that humans do not do well, processing huge amounts of data. There is no training that’s going to get a human to be able to go out and look at a million images in a matter of two minutes, so I think that is where it’s impactful and that is why it’s not a clear “just remove them from here; I’m going to move them on top.”
STAN: I think you’re right on that and I think it has always been easier said than done to take people who have been doing more mundane and transactional work and to make them more strategic. It is a great aspiration but does not always work. And I think as you just highlighted, with the ability to process huge amounts of data, there’s only so much people can do or cannot do in that respect. So I think it will be very interesting to watch, but obviously this is the reality that’s happening and the clients need to understand when and where to adopt and how fast.
About the authors:
Dave has spent the last 30 years helping companies and clients leverage technology to deliver enhanced business solutions. The more recent part of this 30 year road trip has utilized his early-career hands-on experiences to help clients better understand how best to leverage IT technology without being consumed by it. Most recently Dave’s focus has been on robotic process automation (RPA)—a technology approach to displacing client knowledge workers with “virtual workers”—just another method that Dave can use to help clients craft their unique approach to transforming the way they do business.
Stan leads research and thought leadership efforts for KPMG Global Management Consulting, focused on trends, issues, and futures in enterprise services transformation and optimization, the threats and opportunities from market disruptions and disruptive technologies and best practices in responding to and capitalizing on these market trends.