Automating Automation. Readjusting the mindset to process automation.
For some, the title might be a contradiction in terms. How can you automate what is already automated? For others, it might be the wrong choice of words. Surely, you are talking about orchestration, namely scheduling, and integrating automated tasks between complex environments and services! And yes, I am not the first one who is talking about automating the automation. Beyond just shamelessly aiming to grab your attention, the point I am trying to make is that we have to be clear about what we are trying to automate, what the goals are and what progress we have made with automating our processes. So what we need is clarity as to whether we are talking about task automation and predominantly employee-centric productivity gains or whether we rather should be focusing on end-to-end automation for organizations to stay relevant as markets shift toward platform and knowledge economies. If it is the latter, then we need to address how we get to scale and finally get closer to the notion of straight-through-processing that we have been discussing for ages. A critical benchmark here is the capability to run processes autonomously without any agent support if required and not just augmenting agents which is the dominant mode for task automation. Thus, discussing automating the automation is all about how we get to true end-to-end automation, a goal that has become even more urgent with the disruption of Covid-19.
Many of these points were touched upon with refreshing clarity by Citigroup’s Head of Innovation and Productivity, Capital Markets Operations and Technology, Ben Rayner, in a recent webinar by the good folks at HFS. Ben was talking about Citigroup’s journey to transforming their processes and their lessons learned. Much of the journey is summed up by his belief that his function is primarily a change management function even though he runs predominantly a “technology shop”. Yet, at the core of the activities of his team is process improvement where technology levers help to stimulate change. Unsurprisingly, the discussions on RPA some years back had galvanized new waves on process innovations. While RPA had given them good, albeit “isolated”, benefits, Ben reflected that RPA is only automating a part of a process. The challenge for the organization was to provide automation for end-to-end processes. He was sanguine that his organization will always be siloed and will always have multiple platforms that need to be integrated and managed. In recognition of these systemic issues Citigroup started to overlay its process landscape with a single process standard, and a single automation standard. The work on the latter saw them getting deeply involved in Machine Learning and Natural Language Processing as the intent was to integrate ever more heterogeneous data in a near real-time fashion. As Ben put it, those innovations were layered on top of the standardization (and on top of rule-based automation). While he articulated all this much more eloquently than I have, the point that made me sit up in chair was a different, albeit related one. The progress with and the ultimate scale of those deployments of innovative technology are dependent on the ways in which the plethora of changes that are happening to the systems environment underpinning those processes are managed. Examples for such changes include system upgrades, patches, version controls, and similar issues. The activities to scan for those changes are highly manual in nature. While the fixes are often not too complicated, what is much more challenging is to understand what caused the issues in the first instance. According to Ben, nobody has yet found the perfect answer on how to tackle these problems. And in times of Covid-19, these challenges get exacerbated as employees might not have physical access to locations to do those highly manual activities. Against this background, I will discuss the levers of progressing toward end-to-end automation, with self-remediation and autonomous execution as the desired end state. This goes to the heart of what we here at arago are trying to achieve.
Managing automation beyond RPA
Our partner, ChoiceWorx, has addressed many of the change challenges that Ben had called out. I have already covered the details of their strategy in a recent blog interview with the CEO. While the broader industry is obsessing with changes of objects on a screen or changes to information such as the font size on an invoice as part of the RPA discussions, ChoiceWorx is managing the dependencies on the processor, operating system, and application levels that often lead to bot failures. Leveraging our HIRO platform, they are not only monitoring and diagnosing the performance of bots but are also enabling the self-remediation of the underlying cause of failures. Those activities were our bread and butter as we started our automation journey in IT scenarios.
Beyond the glare of the RPA discussions, ChoiceWorx has developed the same approach for end-user support and device management. This is not only an indicator that our HIRO platform is process agnostic as it can automate IT and business processes, but it is also a reminder that IT Automation continues to play a central role in the way organizations manage service delivery, even though the hype around RPA appears to suggest otherwise. The key strategic lever here is that the manual intervention for IT support is reduced to a bare minimum and thus we have true Digital Labour. With that, we are back to the point of process automation. Having the capability to run processes autonomously is the benchmark for effective end-to-end automation. Thus, ChoiceWorx is a compelling example for automating the automation. It is a completely different proposition than what AutomationAnywhere is referring to, while using the same phrase. Their Discovery Bot is recommending tasks to be automated. Yet, those tasks are confined to standardized process flows and do not automate any automation other than RPA. To support organizations beyond productivity gains we need additional approaches that can manage the ever-increasing heterogeneity and variability of digital processes.
Fig1: Arago’s HIRO is automating the automation
Source: Arago 2020
What are the key levers that enable HIRO to automate the automation?
In Figure 1 I have tried to visualize many of those aspects. From a service delivery perspective, there is a continuum of activities from requiring agent support to fully autonomous processes. The more you progress toward the right-hand side, the closer you get to end-to-end automation. Crucially, on a functional level HIRO can be applied to both IT and business processes. Thus, you can not only overcome organizational siloes but you can address the dependencies between IT and business that both Ben and ChoiceWorx have called out. The activities depicted are typically enabled by a plethora of point solutions including automation. However, these solutions are deterministic in nature and require enormous integration efforts. As such, they tend to be barriers to the journey toward end-to-end automation. Therefore, the cogwheels are meant to emphasize that HIRO is managing and optimizing an ecosystem with high levels of interdependency which leads to the ultimate high automation rate. It is exactly this ability to deal with complex interdependency and highly process variability that sets HIRO apart from other approaches.
The key differentiation for HIRO therefore, is the ability to self-remediate and execute processes autonomously. These abilities are essential for integrating with and enhancing other automation approaches. Because the engine must be able to deal with an enormous amount of variability stemming from diverse triggers, alerts, events, etc. as well as changes to the environment. In a nutshell, HIRO is enabling the existing processes to adapt to change exactly because it is not applying a deterministic approach of defining process steps. At the same time, it is important to call out that HIRO is not supplanting all those automation tools and solutions but rather is improving and optimizing them. Thus, the investments of clients get ringfenced instead of becoming obsolete. Without wanting to go into great technical detail, in my view the following three capabilities are crucial to progress toward end-to-automation:
- Self-learning capabilities make HIRO process agnostic: Unlike traditional automation approaches, HIRO is neither pre-optimized for specific processes nor does it follow a deterministic set of process steps to automate processes. Rather, the approach is goal-oriented in that the engine is trying to solve a problem autonomously in the most effective way. In terms of philosophy, it is close to what the good folks at Google DeepMind are trying to do. The problem-solving and consequently the automation is achieved by applying retained digitized employee knowledge whilst leveraging Reinforcement Learning. We refer to this as Knowledge Automation. This allows HIRO to automate even non-standard tasks. Thus, it is a fundamentally different (yet complementary) approach to the journey of Citigroup with a single process and automation standard. But as the infographic shows, what it effectively does is to enable clients to automate IT and business processes with one engine.
- Self-remediation and autonomous execution drive end-to-end automation: In terms of automation, what separates the wheat from the chaff (i.e. the many tools having an “automation” moniker) is the ability to autonomously execute process steps or even complete processes. While the engine must be taught and as such requires a human-in-the-loop, with increasing maturity clients can decide to run processes autonomously. We have successfully demonstrated this in IT Automation and are now applying the same approach to business processes.
- Transfer Learning transforms Digital Labor: If anything, the challenges brought upon by the Covid-19 pandemic underline that we need a new quality of Digital Labor. Because we can no longer assume that human labor is a given to deliver services. So the old Business Continuity playbooks need to be rewritten. The efficiency gains by deterministic approaches are welcome, but they struggle to adapt to changes in the environment. These changes now include a lack of human labor. By blending self-learning (i.e. non-deterministic automation) with a transfer of knowledge from one domain to another, HIRO provides a new quality of Digital Labor. As an example, the knowledge of automating IT requirements provides the knowledge as well as the business logic to help to automate business processes like Finance & Accounting or Supply Chain. In other words, just like task automation will not achieve by itself end-to-automation, Digital Labor must evolve beyond domain-specific knowledge to support end-to-end activities.
Bottom line: Be clear on the goals for automation to cut through the marketing noise
It is not easy to navigate the discussions on Intelligent Automation these days. With the leading RPA vendors gearing up for IPOs, the noise levels will probably get even louder. But to progress with your transformation journey, you need to be clear on the outcomes you are trying to achieve. If your goal is end-to-end automation you should consider complementing your existing automation approaches with automation to automate the traditional approaches. If your goal is to stay relevant on the journey toward the Knowledge Economy, you should evaluate propositions that are not deterministic in nature. Yet, to succeed on your journey, above all else you need an innovation mindset and the ability to drive change. Those are the topics that we would love to discuss with you.
