Not too long ago the idea of using artificial intelligence (AI) would have been looked at as science fiction by nearly any government agency. In June of 2017 at KPMG’s Government Intelligent Automation Forum, many attendees expressed their reservations. Even though the risk of ending up in some sort of Terminator scenario was not deemed likely, the idea of humans losing jobs to machines was at the forefront of executives’ minds.
However, since then, there has been tons of information highlighting the benefits of AI and intelligent automation, and many agencies have started to adopt it. Unfortunately, many of those cases involve applying AI to address a specific problem without a strategy to scale it across the entire agency. Of course, this is a good start, but the benefits could only accrue as automation is used in more and more areas.
Deploying AI and intelligent automation at scale – fully deployed across multiple processes – requires expertise in several skill sets such as workforce and risk management. This requires working with a partner who knows the ins and outs of such a daunting task, and one who has already overcome these hurdles with other government organizations.
KPMG has broken down five key concepts that any program or technology officer will need to consider when piloting, deploying and scaling AI and automation. In their experience with helping more than a dozen government organizations plan and implement AI programs, they developed the following five key concepts from lessons learned. Anyone following these ideas can avoid the most common pitfalls that stymie less experienced organizations.
1.Establish an agency-wide digital strategy
Instead of specific offices adopting AI to meet a specific need, all programs should be involved in creating a plan that will be in line with the agency’s overall mission. In order for AI and automation projects to succeed in production, an agency needs to set clear objectives and goals from the outset.
“The widespread application and potential benefit of this technology for DLA is extraordinary, given the large number of business processes that can be automated with robotic process automation (RPA),” said John Lockwood, RPA program manager in the Defense Logistics Agency (DLA) Information Operations in a DLA article last December. While full implementation of AI across the DLA is years away, they have set clear goals for finding candidate processes for automation and a definitive schedule for execution.
2.Create and communicate the workforce plan
While most government organizations have no plans to lay off workers as a result of their AI and automation deployments, this does not prevent the average employee from having doubts. Organizations need to address this clearly and openly, and emphasize automation’s role in enhancing rather than replacing the work of their staff.
The best way to alleviate any such concerns is to educate workers on how their specific jobs would improve by having certain tedious, antiquated or inefficient tasks made easier through automation. In addition, automating processes can help retain information generated from the years of hard work of soon to be retiring employees.
3. Address governance and policy changes
As one might expect, sweeping adoption and implementation of automation technologies is not generally covered in most organizations’ existing policies. From choosing government-approved technologies to figuring out issues like access control, governance must be considered at the start of any automation strategy.
For example, KPMG worked with an agency through the proof of concept phase, which resulted in a plan to implement more than 20 RPA bots very quickly. Unfortunately, this raised serious security and governance concerns. To counter this, the agency was advised to centralize their automation configurations oversight, which negated the concerns and minimized risk.
4. Evaluate business processes and data
It’s no secret that the government has a vast and ever-expanding sea of data. However, the relative lack of data that is structured and machine-readable can make for a serious hurdle to overcome. At the recent Artificial Intelligence and Intelligent Automation Forum, Michael Conlin, chief data officer of the Department of Defense said, “You can’t feed the algorithms if you don’t have data—solid, clean data in large volumes, well-tagged and well organized. People will tell you that the machine learning algorithms, AI technologies can clean the data for you. Good luck.”
Also, if the processes to be automated are inefficient, the resulting automation will be inefficient. Thus documenting and reviewing all processes prior to automation can first show the necessity to revise or even remove it entirely. Meanwhile, the processes most suitable for automation should bubble to the top.
5. View the CIO as an innovator, not just a technology guardian
Since much of the interest in AI has been generated by departments such as finance, compliance and procurement, the CIO is often looked at as just some sort of glorified gatekeeper. However, their actual job is to integrate technology for the success of the entire organization. Groups need to treat the CIO as an innovator who can aid them with the entire automation experience. The CIO and external providers can also recommend technology solutions that go best with existing systems, as well as follow mission objectives.
KPMG can help any government organization deploy automation using these five core concepts, quickly finding the best places to install the technology, negating any risks and concerns in the planning stages and ensuring that every new implementation fits in with an agency-wide automation strategy. They have the experience and knowledge to help government get its automation programs right from the start.