Agentic AI is moving into the federal government at scale. These systems work autonomously to find and synthesize information. IDC research for Salesforce found that 82% of government organizations have deployed AI agents, and agencies including the Department of Labor, the Air Force, and the Veterans Health Administration are using them to automate customer service, modernize legacy systems, and reduce administrative workloads.
Paul Tatum, executive vice president of global public sector solutions at Salesforce, said AI agents are not chatbots but "autonomous digital workers" that are changing how work gets done and increasing the capacity of overburdened government workers.
"With agentic capabilities, AI becomes more than just a conversational tool; it becomes an execution engine," Tatum said.
How agentic AI works
Agentic AI links humans and agents in complex business tasks with multiple steps, according to Jason Andersen, vice president and principal analyst at Moor Insights & Strategy. Agents take AI models, workflows, and data sets from a wide range of sources and interact with other applications, he said. You can delegate a task to an AI agent, and it will only report back after it accomplishes the goal or gets stuck.
With AI agents, you outline a series of steps and specify the order of actions, Andersen said. You also note an objective, such as delivering a report to a manager or solving a customer problem, and specify timing - like having an agent run reports Monday through Friday. Federal agencies can scale agents up in a cloud environment to add capabilities during peak times.
Where agencies are piloting agentic AI
Customer service is a key use case. AI agents at the Department of Labor and the United States Postal Service can eliminate the hold time customers experience when calling support lines. Andersen said AI agents can take on the entire process of solving a customer problem from beginning to end.
In March, the Department of Labor said it would use Salesforce Agentforce to modernize its National Contact Center. A Department of Labor agent will collect intake details, open formal cases, and escalate cases autonomously to workers when citizens request it.
AI agents also let federal agencies modernize old applications, which have been a hindrance to government operations, Andersen said. He noted that upgrading and migrating this technology is often expensive, and agents can save money on these projects.
The Air Force is using Salesforce's Missionforce National Security to transform operations and modernize mission readiness, Tatum said. The platform connects data, systems, and workflows for airmen and guardians in a single operational view. The Air Force is also piloting Salesforce Agentforce Public Sector to automate complex workflows and sharpen decision-making in the field.
The Veterans Health Administration deployed an agentic operation system incorporating Slack AI to more than 150 VA medical and outpatient centers. Tatum said the platform will save thousands of staff hours, letting front-line workers spend less time on paperwork and more time on patient care. The Department of Veterans Affairs also awarded Salesforce $1.6 billion as part of a three-year Agentic Enterprise License Agreement to deploy AI agents across VA operations.
"AI agents take the paperwork and friction out of government interactions, enhance decision-making by surfacing real-time insights, speed up permitting and licensing, automate administrative work like benefits eligibility checks and case updates, and free up human workers to focus on the highest-priority constituent needs," Tatum said.
These deployments are part of a broader push into AI for Government that spans customer service, national security, and benefits administration.
Guardrails and building a roadmap
In February, the National Institute of Standards and Technology launched an AI Agents Standards Initiative within its Center for AI Standards and Innovation to make agentic AI systems interoperable and more secure. The center, known as CAISI, aims to create an ecosystem of AI standards and protocols with federal partners including the National Science Foundation.
In June, the White House issued an executive order that called out the use of AI agents by bad actors as part of security breaches. That means federal agencies need guardrails to monitor how agents are used. Andersen said the models have reached a level of sophistication where this is a realistic threat, and since the model doesn't govern the agent - it operates independently - agencies need policies to make sure agents behave as intended.
Evaluating agentic AI readiness starts with examining the context and how all of the different data sources feed into the agent, Andersen said. Tatum added that readiness means having a unified, secure data foundation so agents can reason across an entire enterprise.
"The agencies getting the most out of AI made the architectural decision first: Consolidate onto a unified platform, establish a single authoritative view of your data, then layer AI on top," Tatum said. "That's what allows agents to drive real mission outcomes rather than just automating tasks in isolation."
As a first step in building a roadmap, understand the requirements of what you are trying to build, Andersen advised. He said there may not be a hardwired framework or process, which means more testing and more piloting. Tatum advised that a roadmap should include a conversation with workers on what aspects of a workflow feel repetitive or low-value, and recommended beginning with high-value use cases that have clean data and easy-to-measure outcomes.
"Gather feedback, prove the value, then expand," Tatum said. "What starts as a focused deployment becomes a blueprint for the entire organization."
He stressed that human oversight should be part of an agentic AI roadmap as a "feature" rather than a "constraint" to produce agentic AI that's "not just efficient, it's accountable and built to last."
Why this matters for operations professionals
Agentic AI is already deployed at scale in the federal government. The pattern is clear: agencies are using agents to cut administrative work, speed up case handling, and modernize systems that were previously too expensive to upgrade.
Operations teams can apply the same playbook: consolidate data, pilot high-value use cases, keep human oversight in place. AI for Operations starts with the same steps - audit data quality, identify repetitive workflows, run small pilots, then scale.
Your membership also unlocks:
