- Published: October 9, 2019
- Written by Nathaniel Palmer
Over the last decade as we have championed the need for organizational agility and adaptability, we have simultaneously promoted Adaptive Case Management (ACM) from the generations of process automation which came before it. In this context we have most often contrasted “automation” as being both limited and antiquated in the context of modern knowledge work and the inherent benefits of ACM. In this article we will explore how “Intelligent Automation” has emerged as a new discipline for business transformation, combining the traditional capabilities of ACM and Business Process Management (BPM) while extending these beyond what had previously been thought to be the limitations of automation.
This article excerpted from Intelligent Automation: Rules, Relationships and Robots; Designing Strategies and Practical Implementation. Available from Future Strategies Publishing.
Digital Transformation and the New Automation Imperative
According to a recent survey by advisory firm Deloitte & Touche LLP, 95% of CEOs and 97% of corporate board members cited “serious threats and disruptions to their growth prospects in the next two to three years.” The specific threat most frequently cited is the disruptive effect of digital technologies deployed by competitors, and their internal challenge of keeping pace via new technology investments. To put it in trendier parlance, what they fear most is digital disruption. What is the remedy to digital disruption? Digital Transformation – one of the most frequently cited phrases in current business circles. But what does it mean? In the most basic meaning, Digital Transformation is the transformation or “digitizing” of existing processes and operations into a software realization. It is not simply an “app” nor becoming a “dot-com” enterprise, but leveraging digital media (mobile apps, the Internet, smart appliances, et al) to connect with customers, partners, and even employers.
The push for digital transformation and looming threat of digital transformation is old news. We have all seen stories of once dominant businesses who were run out town by digital disruption. Examples include (most famously) Blockbuster being displaced by Netflix. We see around us countless new and innovative digital natives rapidly outpacing erstwhile market leaders. We see firms such as Uber, Facebook, Airbnb, Postmates, and other success stories who ascended from ideas virtually unthinkable a decade ago to multibillion-dollar enterprises today.
While many a CEO may lay awake at night worrying about being “Uberized,” a more pressing issue is how to align digital transformation with corporate governance. How do you avoid the risk of losing control of governance processes, and avoid the risks of security breaches, while ensuring digital access to your products and services by customers? Or enabling partners and employees to engage via digital media? The answer, from a technology investment standpoint is Intelligent Automation, the engine of digital transformation.
Intelligent Automation is the evolutionary step for Business Process Management (BPM) and workflow automation, complementing the use of business rules and process management technology with software robots and artificial intelligence (AI). The critical path to successful digital transformation (with managed risk) is combining information governance with Intelligent Automation.
Digitizing Business Processes
Intelligent automation technologies support interactions with humans, as well as perform work as humans would do. A relatively simple example is an AI-powered “chatbot” able to interact with customers (and increasingly partners and employees) in a way which would otherwise require a human Customer Service Representative. These interactions blur the lines between human and machine. When the interaction via digital media (mobile app or website) it can be difficult or impossible distinguish whether the entity on the other end of the interaction is living or virtual. They easily pass the infamous “Turing Test” (the test developed by Alan Turing in 1950 to assess machine's ability to exhibit intelligent behavior indistinguishable from that of a human.) Yet chatbots are just the thin veneer of Intelligent Automation. Rather, the processes which happen behind the scenes ultimately define digital transformation.
Specifically, a successful digital transformation strategy is one that ties together discrete moments of automation within a more comprehensive, end-to-end process, adhering to rules of corporate and information governance. Supporting this requires a clear model for the separation of concern between the rules of how work is completed and the systems that support it.
In most enterprises the control points for enforcing the rules and policies of corporate governance focus on human beings. They are part of user interface of core business applications. Humans are part of the reporting systems for ensuring compliance with established policies and procedures. Firms focus on the actions of workers (human beings) who apply their knowhow and subjective judgment to perform work.
A chatbot may be able to check the status of an order, or an insurance claim, or even initiate one. But traditionally the high-value work to process that claim or fulfill that order is left to skilled human workers. This is work is often assumed to be “un-automatable” requiring logging in and out of different systems to complete the process (or even a single task). This work typically involves third party systems or otherwise environments which cannot be integrated through a programmatic interface. Instead people do it, from swivel chairs using sticky notes, and as a result the design of the related rules and workflows are based on how the applications were built, rather than the actual objectives of the end-to-end process which span them.
Intelligent Automation allows these existing user interfaces remain intact, enabling software robots perform the same functions as a human user. This allows the existing control points and reporting to remain intact. The work is indistinguishable between human and robot, as the same systems are used. Yet who tells the robot what to do? This is critical role of information governance, ultimately the serving as the lifeblood of digital transformation, by enforcing the same rules that are applied to human workers and ensuring the same level of transparency (including audit trails, records management, and other means for capturing the chain of custody for how sensitive information is handled).
When Automation is Intelligent and When it Isn’t
Investment in new technology is driven foremost by the goals of increasing execution capacity (scalability) and organizational agility. Seeking to “more, faster, and most often” with fewer personal, enterprises prioritize technology investments which can speed time to market, that empower workers to make better informed decisions, as well as to reduce the overhead otherwise required with delivering products and services to market. The ability to adapt and respond according to both new events and consistent with existing rules and policies is critical to organizational agility. Yet this goal is often at odds with automation focused on scalability and repeatability. Current process automation looks a lot the picture below, with a complex set of conveyor belts designed for optimal efficiency and consistency. Industrial engineers designed the ideal routes to move packages in the most efficient way possible, and these pathways are fixed. They do not change or adapt their paths based on what is in the package.
Figure 1: Process automation currently looks a lot like these conveyors, with fixed pathways and process flows designed by architects and engineers, not adaptable to the context of work and business events
The majority of process automation systems currently deployed were designed and built in the same manner as broader business automation initiatives have been for the last few decades. This presented a model of automation rigidly following fixed pathways which are not consistent with the way we work. We do care about what’s in the package. We cannot fully script out in advance the sequence of steps and end-to-end processes without knowing the exact context of any given task we will be performing. For this reason, process automation to date has been limited to repetitive and relatively simplistic process areas. When we combine case management and data-driven intelligence with process automation, we can expand the range of what can be automated or otherwise managed. This combination of capabilities enables Intelligent Automation.
What does Intelligent Automation look like? Using the same metaphor as before, see figure 2 showing one of Amazon's fulfillment centers where its Kiva robots have replaced the fixed conveyor belts. Just as we do in our own work, the robots do care and in fact know what is in the package. Using this awareness of context (what’s in the package and where it’s going) the robots determine the best pathways and placement of products to enable the fastest possible fulfillment process. The robots leverage process, rules and data to define pathways which adapt to the context of work at that moment, just as we need to adapt on successfully complete our work.
Figure 2: Intelligent Automation leverages the efficiency of automated actors with data-driven intelligence that leverage rules and analytics to enable goal-seeking optimization and decision making.
The combination of workflow automation and data-driven machine intelligence supports our ability to manage work while dynamically adapting the steps of a process according to an awareness and understanding of content, data, and business events that unfold. This is the basis of Intelligent Automation, enabling data-driven processes adapting dynamically to the context of the work, delivering the efficiency of automation while leveraging rules and policies to steer the pathway toward the optimal outcome. The process of the case is defined by the underlying policy and rules combined with the information that we gather along the way.
Figure 3: Intelligent Automation as an Integrated Digital Platform
Robots enter the Workforce
Robotic Process Automation (RPA) is one of the fastest growing sectors of business technology, yet one which is often misunderstood. Some refer to RPA as, “It’s just screen-scraping on steroids!” No, it is far more powerful than that. Indeed it is merely a pillar, albeit a critical one, of Intelligent Automation. RPA itself is geared for scale and repetition. It replaces the subjective decision making applied by human beings and lacks the data-driven optimization offered through AI.
Yet while RPA correlates more closely to “Automation” than “Intelligent,” its benefits can be compelling. Leveraging RPA as part of a broad Intelligent Automation approach offers the same opportunity for execution advantage, as well as an equivalent potential for business disruption, as adding physical robots into the enterprise workforce. Unlike solutions whose function is to coordinate and sequence tasks for humans to perform, RPA specifically acts on behalf of humans to perform work – e.g., RPA automates human tasks (manual work) rather than simply machine tasks, as with traditional software automation. Existing user interfaces remain intact, and the software robots perform the same functions just as a human user would do, in passing security credentials as well as entering and/or accessing data from the application it has logged into.
Intelligent Automation bridges the “islands of automation” where humans are the integration points among systems that otherwise cannot communicate. This is work which cannot be automated any other way. By definition it requires logging in and out of different systems to complete the process (or even a single task) and these often third party systems or otherwise environments that cannot be integrated through a programmatic interface, and so they aren’t. Instead humans do it, with swivel chairs and sticky notes and, as a result, the design of the related rules and workflows is based on how the applications are built, rather than the actual objectives of the end-to-end process which span them.
In Figure 3, the Intelligent Automation Platform is presented as a set of core capabilities. This visualization is intended to present the inherent synergy and interplay among sets of capabilities and does not suggest that these are modules within a tightly coupled monolithic architecture. Rather, Intelligent Automation exists as a layered architecture of best-of-breed components which work together yet most often run within their own environments. There is a necessary separation of concern between each layer, allowing for the leverage of best-of-breed components, and increasingly cloud-based services.
RPA by itself most often has no interface. It acts on behalf of the knowledge worker, rather than serving as a core system with which they interact. This underscores the fact that Intelligent Automation is not a category of software but rather a design pattern for leveraging best of breed components for delivering a powerful set of capabilities. At the foundation of any solution for either Intelligent Automation or case management in general is a data layer. This includes an Operational Data Store (ODS) for driving the actions and operations with which the knowledge work is engaged. In addition, there are necessarily one or more Systems of Record (SoR) which stores the data which comprises each case record and its supporting context.
Above the data services record are three distinct but synergistic components which provide the “brains” of Intelligent Automation. These three are RPA combined with a BPM System (BPMS) and Decision Automation package for defining and managing decision logic. On the latter, Decision Automation (or Decision Management) should be understood as more than Business Rules Engine (BRE). The engine is merely the execution, but Decision Management is a relatively new category of software which facilitates the definition and on-going management of rules and policies as distinct artifacts and business assets.
Figure 4: The Intelligent Automation Platform is derived as a layered stack of best-of-breed cots components
The Three R’s of Intelligent Automation
So far, we have discussed in quite detail about the role of Rules and Robots in Intelligent Automation. In many discussions “robots” and “AI” are treated as interchangeable when the reality is that with AI, we typically mean is learning systems; Machine Learning and neural networks (part of the domain of “Probabilistic AI”). In the context of robots, we're talking about something that is very prescriptive. In this context we can think of RPA as example of “Deterministic AI” which driven entirely by instruction sets. RPA doesn’t make independent decisions other than that which is driven by business rules. Rules are by necessity unambiguous. There's no room for nuance or ambiguity with business rules or robotic automation. The “Three R’s” of Intelligent Automation are Rules, Robots – and Relationships.
The third R for Relationships related to how Intelligent Automation shifts the focus toward providing intelligence is understanding how to get data as opposed to having to replicate and store that data locally. Think about how many sources of data are used to present the complete view of a given customer. It is no longer the case that you can expect a single customer data repository. Rather we may have single meta-model capturing the dimensions of that customer, yet the data required to complete an end-to-end process with that customer will inevitably exist on a multitude of systems. Often the majority of required data may be within systems that are outside of the span of our control.
Consider this contradiction – robots and rules require data, which increasingly lays outside of our span of control, and requires specific formatting or transformation to be usable. This is the role of the third R – Relationships. If we know where to look (i.e., if we relationship between events, rules, and data sources) we can use Robots to find data. We can also use AI to validate real-time event and predict the most likely answer when the story is otherwise incomplete. This is where Intelligent Automation may apply as “probabilistic AI” (as distinct from deterministic) to bridge the gap where data is incomplete and thus insufficient to satisfy the requirement for rules and robots.
By knowing meaningful context about data, then applying Machine Learning and other packaged AI capabilities is required for predictive capabilities. Certainly, it isn’t magic, and requires that both business rules and training data are thoroughly validated and unambiguous. Yet both (rules and data) are clearest in hindsight – we see the complete picture once a process is complete. Thus we can leverage Machine Learning to read and learn from historic data, then evaluate in live scenario what the most likely answer is. In this scenario we would typically apply confidence thresholds to determine with the “guessed” data is acceptable or not.
For example, if transactional step as part of an automated process requires a customer’s Social Security Number, which happens to be missing and not otherwise queryable, can use AI to provide likely alternative sources of information. There will likely be various potential candidates and the Intelligent Automation platform can use AI to both find options and evaluate them; selecting the option with the highest level of confidence (e.g., probability) and only allow to path through if it makes a minimum threshold, such as 90% confidence. To the rules or the robot, it is simply the answer they are looking for, yet the intelligence in Intelligent Automation is able to close the gap where there would otherwise be an error or break in the process.
Where to Begin With Intelligent Automation
What is the best starting point for leveraging Intelligent Automation? One strategy is to look first at repetitive human tasks, where users are bogged down performing tedious work, repetitive steps, or otherwise where users are shifting back and forth among different application interfaces as part of the task or process step. These scenarios offer low hanging fruit, yet plans shouldn't be limited to the easy targets. Rather, your strategy should lay the groundwork for horizontal scale, tying together discrete moments of automation within a more comprehensive, end-to-end process. To support this, develop a clear model for the separation of concern between the layers of Intelligent Automation (as illustrated in Figure 4) and specifically that between the capabilities traditionally associated with BPM and RPA. It is worth noting that BPM was never designed to fully replace the work done by human beings, but rather to facilitate that work by assigning tasks, sequencing steps, enforcing rules, and other means of work management. In contrast, RPA in fact is purpose-built specifically to replace work delivered less efficiently and effectively when performed by humans.
Intelligent Automation enables BPM and RPA work in concert for far more efficient and effect coordination of both knowledge work and automated tasks. While the synergy of this combination offers great potential, realizing this value does not happen by default. There is not today an established standard or methodology which prescribes the ideal interplay between BPM and RPA, and indeed some of the greatest pitfalls lie in the poorly-defined separation of concern between the two. For example, one of the common mistakes is to create rules within the RPA definition which are complex, and thus miss the opportunity for separately managing decision logic (business policies and rules) from the procedural logic necessary to the automated task. No RPA platform is designed for decision management, yet a well-architected approach can and should leverage best-of-breed capabilities. As part of a broader digital transformation strategy, we use decision management to ensure consistency of business rules, as well as enable workers to make better informed, data-driven decisions.
Leveraging Decision Automation to Drive Greater Value
Consider this in the context of a use case where process automation (and increasing RPA specifically) is most frequently applied today; the replacement of (typically offshore) manual transaction processing. In this context it is assumed that workers perform relatively repetitive tasks related to matters such as application processing or adjudication. One of the greatest challenges in these scenarios is to ensure workers follow the rules and policy guidelines for how work should be performed, which are enforced via training, work instructions and SOPs, combined with surveillance based Quality Assurance (QA).
Imagine an alternative scenario where users are relieved of subjective decision making (i.e., having to rely on their own interpretation of policies and rules) and instead their work flows through a library of business logic where 100s or 1000s of rules are applied to validate data accuracy, to ensure consistency with policy, and to present a data-driven recommendation for the best action to take next. This provides an objective measure (actual reportable data and analytics) to demonstrate that work is performed according to established policy. It also lowers the training burden, by removing the need to understand exactly what to do at each, while ensuring greater accuracy and consistency, as each and every transaction, process step, and data element is checked automatically (rather than applying QA to only a small sample).
Tackling the End-to-End Process
Expand the aperture on this scenario and imagine BPM doing what it does best by coordinating the end-to-end process, managing the sequencing of steps and state of process as it advances the span of control from one step to the next. Now with the much finer grain definition of how work must be performed, consider that many of the steps which had previously required human intervention can now be performed by software robots, coordinated by the master process, with the instructions provided not by an automation script, but a complete set of rules and policies able to scale to the complexity of your business. There is an immensely powerful set of “digital benefits” to be realized through leveraging automation in a way which enables not only improved work management practices, but also increasing the accuracy, efficiency and quality of work performed by using standard rules, and less reliance subjective judgement, improving data quality and more accurate analytics, while delivering the ability to understand the impact rule and policy changes before they are implemented.
Imagine, for instance, transforming the 1,000s of policy pages and multiple days of training required to support your current knowledge work into a manageable and measureable set of decision models owned and controlled by business stakeholders. How about gaining rich new source of analytics and audit data based on actual decisions made and actions taken, rather than actual surveillance based Quality Assurance? Intelligent automation offers the ability to integrate processes, rather than systems and applications, to deliver closer to holistic or comprehensive automation of work rather requiring (far more expensive) humans to perform this work manually. This provides objective measure (actual reportable data and analytics) to demonstrate that work performed according to established policy. It also lowers the training burden, by removing the need to understand exactly what to do at each step, while ensuring greater accuracy and consistency. Rather than a “black box” of backend automation, each transaction, process step, and data element is checked automatically against business rules.
Expand further the aperture on this scenario, consider the role of traditional BPM coordinating the end-to-end process, managing the sequencing of steps and state of process as it advances the span of control from one step to the next. Now with the much finer grain definition of how work must be performed, leveraging the policies and rules defined as part of business logic, many of the steps which had previously required human intervention can now be performed by “intelligent” robots. Yet these robots aren’t smart, per se. This is not AI run amok. Rather the software robots are held to the same compliance rules and reporting standards otherwise defined for human workers, but digitized as part of an transformation strategy. Over time, the scope of the scope of this automation can grow to encompass an increasing number of erstwhile human tasks, as performance data are captured and more is understood about how the work should be performed. This is the promise of Intelligence Automation; expanding the efficiency of automation while delivering greater transparency and policy compliance. This is why the value proposition of Intelligent Automation can be so compelling. It is the ability to integrate process, rather than systems and applications, to deliver closer to holistic or comprehensive automation of work rather requiring (far more expensive) humans to perform this work manually.