Ushering in a new era of work with RPA and AI

Government is ushering in a new era of work, using automation and artificial intelligence to help the federal workforce achieve higher levels of productivity and decision-making.

Over the past two years, agencies have focused on shifting the workforce to “high-value” work -- a key goal of the President’s Management Agenda -- by taking advantage of robotic process automation and other technologies to reduce error, improve compliance and eliminate repetitive administrative tasks.

Although RPA is a useful IT capability that allows agencies to eliminate low-value, mundane, transactional work, it can only make simple decisions. By adding AI to the equation, agencies can accelerate the ability of RPA to complete a multitude of tasks at once.  This can be particularly helpful when analyzing large swaths of data, enabling decision-makers to meet goals more efficiently and effectively.

The combination of these two technologies has delivered more real, tangible results that can be actively applied to digital solutions for civilian and defense agencies than either technology could do individually.

RPA, which provides software bots to automate high-volume, repeatable tasks within legacy processes and applications, has opened opportunities to massively transform government operations. “Current RPA programs operating within agencies are achieving roughly five hours of workload elimination per employee,” according to the RPA Program Playbook, published earlier this year by the Federal RPA Community of Practice.

The Playbook continues: “If the government deployed RPA at scale and achieved only 20 hours of workload elimination per employee, the net capacity gained would be worth $3 billion -- and that is only scratching the surface.”

RPA, a building block for AI

Many agencies across the federal government have initiated RPA programs to automate tasks of varying complexity across multiple functional areas including finance, acquisition, IT, human resources, security and mission assurance. Popular uses of RPA include data entry, data reconciliation, spreadsheet manipulation, systems integration, automated data reporting, analytics, customer outreach and communications.

In 2019, the Food and Drug Administration’s Center for Drug Evaluation and Research reported it had seven RPA projects in development, including one that automated drug intake forms and freed up the pharmaceutical and medical staff for the agency’s core science mission. Last year, the Defense Logistics Agency completed a first-of-its-kind proof of concept in government that allowed unattended bots to work around the clock. DLA recently reported it has saved more than 200,000 labor-hours with the 82 RPA bots it launched in the past year, CIO George Duchak said during an AFCEA DC virtual event in May. In fact, using basic bots is the first step in the agency’s AI journey, he said.

RPA is transformative because it establishes the building blocks for AI in terms of IT infrastructure and task standardization, the Playbook notes. If RPA is effectively deployed, machine learning (ML) and intelligent automation are only a few, manageable steps away.

RPA/AI use case: Transaction matching, fraud prevention

Applying AI/ML along with RPA provides opportunities for financial management offices to address areas such as transaction matching, fraud prevention and anomaly detection.

For example, large financial management offices struggle to resolve and match hundreds of thousands of transactions, many of which require significant manual effort. An RPA solution can automatically access data from various financial management systems and process transactions without human intervention, but it will fall short when data variances exceed tolerances for matching data and documents and will result in unmatched transactions. The addition of an AI/ML capability would accelerate the handling and processing of data and associated actions, including matching financial transactions or identifying fraud.

If there is an error in the data on a particular transaction, for example, an automated system might not be able to match the transactions with confidence. However, a ML platform could train models to rapidly examine the correlation between historical and current transactions. It could help identify potential matching or irregular behavior based on transactions that might have erroneously mismatched fields such as different dates or name variants. This capability would accelerate the review process and preserve humans for the most important activities.

To be effective, a ML platform must adhere to open standards and offer an extensible set of tools that enable end-to-end data science and RPA development in a rapid, scalable and sustainable manner. This will allow agencies to innovate further as their data maturity and AI efforts improve.

As the power of AI grows with each new use case, so too do the misconceptions surrounding the technology, particularly the erroneous idea that AI will replace human workers, impacting their livelihood as the technology overtakes their job.

RPA has proved it can automate the manual, repetitive, low-value tasks that often drive worker dissatisfaction. The use of AI should enhance workforce efficiency by deferring boring, time-consuming tasks to computers, allowing humans to then make better, more informed decisions based on proven, trusted data that they did not have to take the time to analyze. By implementing the proper change management and communication strategy, agencies can help their employees see RPA and AI as a path to more meaningful, mission-aligned work.