Best listening experience is on Chrome, Firefox or Safari. Subscribe to Federal Drive’s daily audio interviews on Apple Podcasts or PodcastOne.
From chatbots in call centers, to predictive maintenance of military vehicles, to a more efficient federal workforce, agencies have launched a wide array of automation pilots.
Following the release of an AI playbook from the American Council for Technology-Industry Advisory Council, and less than a year after agencies joined forces to create both an artificial intelligence and robotic process automation community of practice, agency officials have expressed enthusiasm for this emerging technology, but agree that more steps are needed to scale up those applications.
Even in cases where automation has improved operations, getting reliable data on the return on investment can be a challenge.
Rachael Martin, the mission chief for intelligent business automation in the augmentation and analytics division of the Defense Department’s Joint AI Center, said DoD has seen better luck with calculating the impact of automating financial management workflows, for example, but have had a harder time quantifying the human capital impact.
“When you have a military officer assigned to an admin office and maybe they’re stuck doing paperwork, allowing them to get away from that paperwork to focus on more warfighter initiatives is both a morale issue and a better resource management issue from a department perspective,” Martin said Wednesday at an ACT-IAC conference. “Sometimes ROI is not tangible, and might be something that is harder to quantify in a way that is understandable for a business manager.”
The JAIC, Martin said, currently has about 70 civilian staff and about 30-40 contract staff, but the agency expects to double its civilian workforce by the end of fiscal 2021.
When the JAIC was stood up in June 2018, it began its work on immediate issues like predictive maintenance of military vehicles and cybersecurity. Since then, Martin said the agency has focused on projects like warfighter health and intelligent business automation, which includes using AI to make DoD processes more efficient and accurate.
Harry Lee, the assistant commissioner of GSA’s Technology Transformation Service, said the rollout of AI and intelligent automation has allowed federal employees to focus more of their time and attention on core responsibilities, and will enable a “more personalized experience” when it comes to overhauling customer experience.
“AI and IA are about doing things in new ways — reaching more people, executing at greater scale speed and precision,” Lee said.
GSA has now deployed 33 RPA bots and has shared most of those on Github and with a governmentwide RPA community of practice. These bots include an accounts payable bot that automatically sends email notifications for outstanding invoices, as well as a bot that pulls information on micro-purchases to create a purchase card log.
But beyond workload reduction, Lee said automation, if used correctly, could reduce human error, increase compliance and allow agencies to act on a growing volume of data.
“Data continues to grow exponentially … I don’t see that changing anytime soon. And to a large extent, the success that we have with data will be dependent upon our ability to use AI to manage the data,” Lee said.
While predictive AI models can help agencies make better sense of their data, and empower leaders to make decisions, Kris Rowley, GSA’s chief data officer, said the work of maintaining those models will require additional effort.
“As soon as you’ve developed a model and you start to make predictions, you’re starting to incur more information and more outcome data and more variables — more ways to bring more data into this environment to do better predictability,” Rowley said. “There is a never-ending churn of work that needs to be done to manage and maintain models.”
Ed Burrows, GSA’s former lead on RPA, now the vice president of intelligence solutions at Brillient Corporation, said agencies have seen a significant return on their investment. His former agency, for example, has saved at least 7,000 labor hours through its RPA bots.
More importantly, most RPA tools, he said, are built as low-code or no-code software, making its implementation more of a workforce management conversation than a technology one.
“The developers aren’t writing lines of code. They’re putting together building blocks into workflows. It’s easier to develop. More people can do it who don’t have IT backgrounds,” Burrows said.
While easy to get started, Burrows said GSA’s success with RPA comes with identifying work processes that are good candidates to automate.
Supporting partners include Presidential Innovation Fellows, GSA’s digital consulting service 18F and the public-private Centers of Excellence now built out at five agencies.
Support also comes from governmentwide communities of practice. GSA stood up its RPA community of practice in April, and now has more than 750 members from more than 50 agencies.
Last week, the RPA Community of Practice released a program playbook for future implementation. The document offers advice for agencies just getting started with RPA, like building a strong partnership between program offices and chief information officers.
According to the playbook, RPA programs have eliminated about five hours of workload per employee. If these programs scaled up to cover 20 hours of workload per employee, GSA Chief Financial Officer Gerard Badorrek, the community’s chair, said the federal government would save more than $3 billion dollars a year.
“Additional collaboration between agencies can help accelerate RPA adoption governmentwide, as agencies are currently wrestling with common implementation challenges in a vacuum,” Badorrek wrote in the playbook.
Meanwhile, Federal Chief Information Officer Suzette Kent last October launched an AI community of practice with more than 400 members from 26 agencies.