How to create the best RPA architecture

Beauty of layered designs

David Brain

Despite the differences in RPA tools, most businesses that deploy automation use a layered design to conduct complex processes. This approach separates the logic and functionalities of a project into different parts with the goal of making the project easier to understand and develop automation. Businesses use a layered design model to allocate roles for both RPA software and team members, creating a clear distinction between the processes that will be automated and the ones that require human intervention. This model also makes it easier for team members to focus on the sections and parts that play up their individual strengths.

Phil Fersht

There's more to layered designs than work delegation; in some instances, businesses will add subprocesses to their layers, which divide individual processes into a hierarchy of smaller components. For example, a larger process in a layered design might be employee onboarding automation, and the subprocess of that would be creating an account for the employee in the expenses system, a task that RPA could very easily and effectively manage.

While this approach to RPA architecture might seem time-intensive -- particularly for complex projects -- breaking processes down into various layers and subcomponents is essential to optimizing both the deployment and maintenance of automation. The more defined each layer is, the more efficient RPA will be in completing a full project -- and the easier it will be for team members to understand where they fit in.

Assisted automation vs. unassisted automation

RPA tools offer two modes of deployment: assisted automation and unassisted automation. Each model has its benefits and limitations, and before an organization formulates its RPA architecture, it is critical to determine which model best meets the organization's needs.

Assisted automation is when an RPA tool automates other applications running on the employee's desktop. It requires a user to trigger the automated steps of a process and is most frequently used when helping people complete complex processes. The benefits of assisted automation include reduced handling times for projects, increased cost-effectiveness, and improved customer and employee experiences. The major drawback to assisted automation is that any inconsistencies on desktop settings -- such as changing graphics, display settings resolutions and so on -- can cause the RPA to fail, and sometimes the desktop is locked when the automated steps are being performed.

Unassisted automation, on the other hand, does not require a human agent. In this instance, the RPA software performs on its own, only alerting employees when something goes wrong. With unassisted automation, the RPA software will work 24/7 -- an ideal scenario for optimizing an entire process. Unfortunately, this does present a significant challenge: In order to work, unassisted automation requires structured information and clearly defined rules, which may not be available in every project use case.

With whichever model organizations decide upon, it's important to not rule out vendors just because their product doesn't support both models; in some cases, more than one tool across an enterprise is the most beneficial approach.

RPA control center

With design and deployments determined, the next step in creating the optimal RPA architecture is ensuring the tool has a fully functional control center with key features related to error handling, process analytics and resource allocation. The control center acts as the central interface from which all process commands are issued. From this center, administrators are provided with the operational agility to properly launch, maintain and upgrade their RPA systems. This makes the management of tasks like resource allocation and project designation more efficient across the enterprise and helps provide a holistic view into how the RPA system is performing.

To that end, organizations must review the features of the RPA control center. Error handling, in particular, is the key to guaranteeing that processes are running at their full potential. While RPA tools might have processes to alert employees when an error occurs, a more comprehensive system will keep a detailed record of each instance, making it easier for teams to track down problems in less time.

Breaking processes down into various layers and subcomponents is essential to optimizing both the deployment and maintenance of automation.

Another key feature of an RPA control center is process analytics, which monitors the real-time status of processes to confirm that the RPA system is performing as expected. The analytical component of a control center is essential to properly auditing an automation tool so that improvements can be made.

RPA control centers should also have the ability to manage resource allocation, allowing administrations to easily control which tasks each machine will perform. The most effective RPA tools are the result of proper resource allocation driven by data derived from the process analytics feature. When used in tandem, these features allow for the most impactful administration of automation tools and give organizations the ability to fully utilize the power of RPA.

While there is a variety of RPA tools available today, deciding which one is best is not the first question enterprises should ask. They must first consider what the RPA architecture for the deployment will look like by:

• creating layered processes;
• identifying if the project requires assisted or unassisted automation; and
• verifying that the RPA control center has all the necessary components.

It's the only way to ensure an RPA deployment is going to be successful from the start.

Editor's note: This is the first article in an eight-part series by David Brain, COO at Symphony Ventures, a consulting and managed services firm, and Phil Fersht, CEO and chief analyst at analyst firm HfS Research.