When it comes to government deployments of automation and AI/ML, the spotlight often shines brightest on high profile applications – whether it’s augmenting soldier performance on the battlefield, driving intelligence gathering or modernizing the delivery of citizen services.
For that reason, it’s easy to miss a transformation underway in the “back offices” of Defense Department and civilian agencies, where adoption of Robotic Process Automation is soaring.
RPA technology instructs a virtual software robot to mimic human actions in order to automate mundane, repetitive, manual-heavy tasks. Ideal automation candidates include processes that are highly standardized with electronic data inputs, such as streamlining data collection and processing, document management, identity verification and automatically responding to information requests from citizens.
A 2021 survey underwritten by UiPath found that 6 in 10 federal respondents — and half of state respondents — view RPA as a building block to harnessing artificial intelligence and machine learning, by accelerating data gathering and improving data quality. Why? For starters, RPA has the lowest cost to entry in the Intelligent Automation space and provides high rates of returns when implemented correctly.
RPA and the use of intelligent bots are proving critical for agencies seeking to automate time-consuming manual processes required to run a government agency. For agencies evaluating RPA, it is critical to first identify which processes are best suited for automation – a decision in large part driven by the impact on the amount of time, money, and frustration saved with a new solution.
This article will detail how agencies at any level can determine the best applications to automate via RPA, as well as the key benefits RPA delivers.
Strategies to evaluate RPA candidates
When evaluating where RPA can yield the greatest operational and people impact, it is key to work through the backlog of candidate processes. In other words, before RPA development even begins, conduct thorough analysis, improvement, and standardization of potential automations. Selecting the right candidates should not occur in a vacuum; it requires understanding the overall mission, the business environment, and the IT systems. Given those considerations, the five strategies below can guide decision makers towards the most effective candidates.
Is the process manual and repetitive?
The work best-suited for RPA consists of repetitive, manual processes that require the attention of an employee and involve following similar steps for every new piece of input. Processes that currently require a person’s involvement or oversight, such as data entry or processing personnel files, stand to benefit the most from RPA, as automation of the process frees up employee time for more meaningful tasks.
Is the Process Frequent or Burdensome?
Some operational processes need to occur monthly, some daily, and others on-demand from thousands or millions of connected users. This consideration should be weighed against the burden incurred by the process.
Let’s look at an end-of-week report that takes four hours of an agency employee’s time to create each week, and compare that to year-end reports that take twenty hours to complete. Although the year-end reports are a more onerous task, the end-of-week reports take place far more frequently, making them a higher-priority candidate for RPA. Automating the simpler, but more frequent process saves two hundred hours compared to the twenty consumed by the “bigger” task.
Does the Process Use Rules or Templates?
AI-powered RPA solutions can be capable of accomplishing complex decision-making, but the processes that follow an established set of rules or instructions make for the most efficient use of intelligent automation. Even if the rules that guide a process are extremely complex — with branching paths of recommended next steps based on several different criteria for each input — a computer can internalize these rules far more efficiently than a human.
Tasks that require some level of subjective judgment, on the other hand, are far more challenging to adequately approximate with a set of hard-coded rules. Templatized processes make for much more straightforward RPA implementations, as AI can learn to run complicated analyses of any input given the proper template through which to make sense of the data.
Does The Process Handle Standard Inputs and Outputs?
Any process that takes in a standard kind of data, such as a document, PDF, or spreadsheet, could be replicated by RPA trained to handle a given consistent data format. Processes that accommodate several different formats, such as a mix of emails, paper receipts, and video recording require far more investment to automate. A standard output is also important - generating a file, noting information in a database, sending out emails or updates on a network, etc. Scanned physical documents can be transformed through Natural Language Processing can or Optical Character Recognition through automated processes to turn these documents into usable data.
How Many Business Applications Are Involved?
The fewer business applications involved, the more efficient RPA can be out of the gate, as the AI only has to understand the inputs and outputs of one program, and can interface with it directly. More advanced implementations can connect several applications and enable them to “speak to” each other, but this level of investment is most useful where ROI is predicted to be substantial.
RPA benefits today and tomorrow
At a micro level, an illustrative RPA use case can be found with The Air Force Installation and Mission Support Center, which faced significant challenges processing FOIA requests and replying with notification letters. Managers were overwhelmed with a backlog of hundreds of cases that took their time away from higher value tasks.
The FOIA process caused significant resource leakage due to the procrastinating nature of the existing process, thus an optimal application for RPA to eliminate the error prone process and realign current resources strategically. By replacing tedious manual tasks with automation, users could narrow their time investment to making critical decisions and reviewing the work the intelligent “bots” performed.
The result: A backlog reduction of 30% and a significant bump in fee generation on Requests Answered in Time. It also decreased the processing time by 88% and increased accuracy to 99.9%. The RPA implementation drove an overall reduction in employee workload and 2034 hours of Government FTE were reallocated for better use of time.
What’s next for RPA? As more agencies experience tangible results there will be a desire to continue the journey to more intelligent automation. RPA is already evolving beyond baseline, rules-based chatbots often used for customer and citizen support. Applications today are leveraging Intelligent Automation (IA), which Brookings defines as “...a type of RPA that includes AI, ML, or natural language processing (NLP). When applied correctly, RPA and intelligent automation make for fewer overworked employees, more accurate and speedy internal processes, and a more capable organization in the long run.
Mark Hogenmiller is Chief Transformation Officer at Aeyon, a provider of management consulting and data analytics services to the federal government.
Have an Opinion?
This article is an Op-Ed and the opinions expressed are those of the author. If you would like to respond, or have an editorial of your own you would like to submit, please email C4ISRNET Senior Managing Editor Cary O’Reilly.