Virginia builds AI-, RPA-as-a-service offering for state agencies

The Virginia Information Technologies Agency (VITA) is working to offer artificial intelligence as a service to the state’s 65 agencies.

VITA is rolling out the Digital Customer Experience, an AI service that’s essentially a basic chatbot that state agencies can implement into user-facing functions. For more complex AI and machine learning solutions, VITA will provide the country’s first public-sector AI as a service.

“What we wanted to do was design a service that avoids the common AI pitfalls,” said Jon Ozovek, chief operating officer at VITA. “AI at this point in time is highly predicated on data, so you have 65 different agencies with 65 different levels of understanding of business processes, 65 different levels of datasets, 65 different levels of data security. That’s a massive problem.”

Al as a service will look like this, he explained: VITA will conduct an intake process for state agencies that want to implement AI to get a solid understanding of the business process it would apply to, the datasets involved and the data maturity. That will help determine where AI can help the most and bring the most data, Ozovek said.

“A business case is associated with one or more tools, which is associated with one or more outcomes, and that allows us to target the service based on our intake process to a very acute set of tools and set of outcomes,” he said. “That’s how we’re able to give them a good idea of what kind of outcome they’re going to achieve from this investment.”

An end-to-end service, VITA will offer proactive monitoring of algorithm drift, in which algorithms become less effective over time, and adjust for performance. It will also monitor cybersecurity, assist with setup and training and establish an AI center of excellence (COE) with standardized outcome and metrics gathering in addition to implementation templates for some AI and ML tools. That means when VITA identifies certain factors in the intake process that can be improved with AI, it can apply a template to speed implementation.

Right now, VITA is doing use-case analysis on processes in transportation, health and public-safety, and it plans to target services that address resource allocation, predictive analytics and forecasting, real-time operations management, decision support and proactive risk management and mitigation, Ozovek said.

Last summer VITA launched the first public-sector robotic process automation as a service offering through UiPath. The goal is systemizing successful implementation of RPA, which allows for the development of bots that can automate many organizational processes. So far, almost 10 agencies use RPA as a service, with another five ready to do so.

The Virginia Department of Health, for example, has used it to analyze and aggregate the thousands of lab reports it received during the COVID-19 pandemic.

“The amount of data availability would be the metric you’d use to measure the process and the throughput of the process,” Ozovek said. “Due to RPA in that particular process and meeting the needs of the pandemic, the data availability increased by 7,000%.”

To address the state’s unemployment insurance response during the health crisis, the Virginia Employment Commission enlists RPA as a service. It can review 120 applications in 24 minutes, whereas the task took a human about 10 hours. That also helped reduce the backlog of claims by about 40% and allowed VEC to reassign workers to more complex tasks.

Another user is VITA itself, which deployed RPA for its supplier invoice auditing process – a highly manual process of auditing and reconciling that ensures the agency’s fiscal responsibility.

“It took us about 16 hours, give or take, to implement the bot that automated that process, but we’re on track to save in the high hundreds of manual hours a year across the board just within my finance department,” Ozovek said.

RPA as a service won’t be a fit for all agencies, he said, adding that his goal is to get RPA to 60% of them. “If I extrapolate the outcomes that would be achieved by doing that for 20, 30, 40 agencies, that’s going to provide immense cost savings, an immense amount of productivity improvements, an immense amount of waste reductions for the commonwealth as a whole,” Ozovek said.

VITA set up a COE for RPA as a service, too. It captures lessons learned from all deployments and features standardized bot development. That way, when an agency inquires about a bot that’s already been designed, VITA can stand it up without having to do new coding.

“We want to have that mind space within our customers’ perception that we are that strategic partner and adviser that they proactively go to when they have a problem or a need solved through technology. How you do that is you establish a pattern of execution of services like this that add value to the bottom line,” Ozovek said.

Still in the pilot phase, VITA had a goal of completing service scoping, introduction and delivery of RPA as a service within 100 days, and is on target to hit that in December, he added. AI as a service will follow next year.