How agencies can use AI to automate and augment operations
Connecting state and local government leaders
Artificial intelligence is ideal for solving comparatively mundane but critical and widespread problems like reducing call wait times and case backlogs.
As artificial intelligence technology becomes more mainstream, federal agencies are beginning to understand how it can be used to redesign mission capabilities and are already planning for this new wave of innovation. According to recent research, over two-thirds of federal agencies plan to make investments in AI technologies in the next year. Additionally, 82 percent of federal executives agreed that within the next two years “AI will work next to humans in my organization, as a co-worker, collaborator, and trusted advisor.”
Far from some futuristic, robotic vision, AI is ideal for solving comparatively mundane but critical and widespread problems like reducing call wait times and case backlogs. It does this in two ways: augmenting human intelligence and automating routine tasks. The combination of these two capabilities allows knowledge workers to spend more time problem solving.
One way to evaluate AI’s immediate potential for federal agencies is to analyze current operations using two criteria: the complexity of the underlying data and the predictability and repeatability of specific tasks or processes. With this information, agencies can identify areas where AI can be used to either automate or augment specific work functions.
For instance, automated solutions are a good fit for routine, predictable and rule-based tasks, such as reformatting handwritten information into electronic text because the verification criteria is repeatable. More ad hoc and unpredictable processes requiring the expert judgment of an experienced benefits case manager or intelligence analyst are better fits for augmentation, because decisions are often contextual.
Now think about the data that support these processes. If it’s structured and machine-readable, it’s more likely to be a candidate for automation. If the data is unstructured and includes speech, social media, video or sensor data, it will probably require interpretation and will be best used to augment human capabilities.
Mapping the complexity of data and processing involved along a simple axis (see insert) can help agencies identify tasks suited automation or augmentation.
It is important to note that different aspects of the same job may be candidates for either automation or augmentation.
AI solutions are already at work in national security and cybersecurity, detecting patterns of anomalous behavior and predicting where bad actors might strike. These solutions have also improved customer experience by reducing call wait times and accelerating case processing through smarter triaging of citizen’s application data.
Using commercial cloud services, Los Angeles was able to quickly create a chatbot -- City Hall Internet Personality, or Chip for short -- to answer questions 24/7. Initially designed to help the city’s 100,000 businesses navigate government services more effectively, Chip has reduced email inquiries requiring manual responses by more than half.
When a federal agency needed to augment case reviewers' ability to quickly verify claimants' information, it used machine learning and predictive analytics to streamline and shorten the evidence gathering process for adjudicating claims. Leveraging the agency’s existing data warehouse, the prototype solution used models to “learn” what evidence is typically required based on historical claims and then generated a recommended evidence list at the onset of each claims review process, eliminating time-consuming and frustrating back-and-forth communication with the claimant
We are still in the early stages of the AI revolution, but the pace of technological change is accelerating. Using tools like the AI task mapping framework above, agencies can understand which solutions are appropriate for a specific task or process and determine where they could have the greatest impact. In parallel, agencies must also invest in change management and communication to demystify AI and help staff understand how they can leverage AI to perform their roles more effectively.