When we think of artificial intelligence, we often imagine something torn from the pages of a science fiction novel. There’s truth to that. But agencies also use AI in less conspicuous ways — to accelerate backend tasks and secure IT networks, for instance — and constituents often have no idea when AI makes decisions about their lives.
According to Deloitte, 60% of current government investments in AI and data analytics are intended to directly impact real-time operational decisions and outcomes by 2024. Seventy-nine percent of government decision-makers are confident in AI’s ability to improve bureaucratic efficiency; 61% believe that AI is at least “moderately functional” in their organizations.
If your agency is not burdened with legacy IT, it’s more likely to embrace AI than your more outdated counterparts — and even likelier if your workforce is tech-savvy. The defense and intelligence sectors are especially enthusiastic about AI.
Below are general categories into which government AI programs fall, along with some real-life examples.
Task Automation
Intelligent automation — which combines AI with workflow automation and robotic process automation — is a powerful way to reduce dependence on manual systems, improve data accuracy and give employees time for more satisfying work. One AI subset, computer vision, can digitize paper documents, while another, natural language processing (NLP), can search digitally submitted forms and unstructured data, such as emails and video files, for critical insights.
The Social Security Administration relies on AI to process disability claims, which has reduced a backlog of pending submissions. The Defense Department’s (DoD) Deep Exploration and Filtering of Text (DEFT) program uses NLP to draw meaningful conclusions from sources including online public comments, social media chats and field agents’ narrative reports. And in Texas, state officials use NLP to read constituent emails and forward them to the correct departments.
Service Delivery
AI helps not only government entities, but also the people they serve. AI-driven self-service portals and “bots” make it easier and more efficient for people to access government data, especially when the agency in question is large or staffing is light. For example, the Internal Revenue Service uses voice bots in both English and Spanish to answer callers’ tax return questions.
The Department of Homeland Security has Emma, a virtual assistant that answers queries and guides about 1 million visitors through the department’s website each month. And the city of Phoenix developed myPHX311, which is both a mobile app and an online portal where an AI chatbot helps residents contact key departments and weigh in on local activity.
Data Analysis
AI can collect and analyze data much faster than people can, leading to better-informed, timelier and more cost effective decisions. One federal example is the Centers for Medicare and Medicaid Services, which uses AI in dozens of ways to analyze health care claims data, identify possible fraud and waste, and offer pattern-ofcare insights.
And local governments, including Washington, D.C., often use automated decision-making systems to aid or replace potentially life-altering human judgments. The systems decide about public benefit eligibility, policing, housing, health care, immigration, education, child welfare and other concerns.
Predictive Analytics
Machine learning algorithms can use massive amounts of data to predict a host of things: addiction trends, military strategy, local real estate values and natural disaster risk, to name a few. At the U.S. Department of Energy, a Transportation State Estimation Capability Tool estimates real-time, street-level vehicle activity to ease traffic flows, relying on data from Uber drivers and others.
Taking a page from private airlines, the U.S. Air Force now uses ML to predict when its aircraft are likelier to break so officials can schedule preventative maintenance. At the local level, the Los Angeles Police Department’s AI models predict crime twice as accurately as trained analysts do, which allows the department to deploy police officers where they’re needed most.
Cybersecurity
Traditional cyber defense tools detect network intrusions 200 days — more than half a year — after they occur. But AI-powered options quickly sort through masses of security data to set priorities for security alerts and address the most urgent concerns first. AI solutions scan network activity, set benchmarks for normal behavior, run advanced behavioral analytics, find cyberattack patterns, prepare countermeasures and automate incident response.
For instance, DoD uses AI automation to protect the many commercial apps the department uses worldwide. And AI and ML help the U.S. Navy overcome the challenges of a complex, siloed IT environment and protect unclassified, secret and top-secret data, among other materials and devices.
This article appears in our new guide, “AI: A Crash Course.” To read more about how AI can (and will) change your work, download it here: