The news
The Transportation Security Administration (TSA) is using an AI agent called Ace, built with Salesforce (CRM), to answer routine traveler questions about airport security, Salesforce said in an announcement dated September 14, 2026. According to the company, Ace handles about 100,000 routine conversations a month and resolves 96% of them without passing them to a person.
Ace is available on tsa.gov and in the myTSA mobile app. Nextgov/FCW reported on September 15 that the agent draws on the TSA website and more than 550 knowledge articles, and that it went live during the summer 2026 travel season. Questions that are too complex are handed to a human TSA agent along with an AI-generated summary of the case.
Salesforce said Ace is built on its Public Sector Solutions product, and its announcement describes the deployment as using Agentforce, Salesforce's AI agent platform.
The company also published cost figures. Salesforce said Ace reduced the cost per consumer interaction by more than 90%, from more than $4 previously. The company said the TSA had needed the equivalent of 12 full-time employees to handle excess workloads, and that Ace takes on that routine work and backlog.
Beyond the public-facing chat, Salesforce said related automation is projected to handle 70,000 internal data-sharing transactions a year, saving 11,660 staff hours annually and producing $11 million in labor-based value over three years. Those are projections from Salesforce, not audited results.
Niki French, Customer Experience Branch Manager at the TSA, said in Salesforce's announcement that easy access to accurate information can help alleviate stress for travelers.
The numbers
- Routine conversations handled per month
- About 100,000 (Salesforce)
- Resolved without human escalation
- 96% (Salesforce)
- Cut in cost per interaction
- Over 90% (Salesforce)
- Knowledge articles Ace draws on
- More than 550 (Nextgov/FCW)
- Projected labor-based value
- $11 million over three years (Salesforce)
- Projected staff hours saved
- 11,660 a year (Salesforce)
Why CEOs should care
For CIOs and customer service leaders, Ace is a useful benchmark because the numbers are specific: monthly volume, resolution rate and cost per interaction. When vendors pitch AI agents for your own support desk, ask for the same three figures from a live deployment, plus how 'resolved' is defined. A question closed by the agent is not always a question answered correctly.
For CFOs, note what the savings are made of. Salesforce frames part of the value as taking on backlog work that needed the equivalent of 12 full-time employees and $11 million in projected labor-based value over three years. Those are vendor projections. Before building a business case on similar claims, ask whether the savings come from cut headcount, reassigned staff or avoided hiring, and what the agent itself costs to run.
For boards and risk officers, the design detail that matters is the handoff. Ace answers only from the TSA website and a fixed set of knowledge articles, and passes harder cases to a person with a summary. Keeping an agent tied to approved content, with a clear route to a human, is a sensible template for any customer-facing deployment in a regulated setting.
The bigger picture
Salesforce has been pushing hard into federal work. On September 29, 2026, Tech CEO Daily reported on a Department of Veterans Affairs agreement for an agentic AI enterprise license with a $1.6 billion ceiling. Salesforce also said in its TSA announcement that nearly 3 million airline passengers fly through U.S. airports each day, a sign of the volume an agency-facing agent can face. Published results such as these give Salesforce a reference case as it competes for more public-sector contracts.
What’s next
Salesforce said Ace is part of a broader modernization effort run by the TSA's Customer Experience Branch. Watch for whether the TSA extends the agent to more channels or tasks, and whether other agencies publish comparable resolution and cost data from their own deployments.
What “Fact-checked” means
Fact-checking means testing a story’s facts against the evidence before it is published. This story went through at least two separate checks before this version was published.
- What we checked
- Its names, figures, dates, job titles, quotes and who said what were checked against the story’s sources, including its main source where it could be opened. The headline was checked for accuracy and overstatement.
- How
- A first check reviewed the whole story. If it passed, a second, skeptical check went back to the sources to look for mistakes in the most important facts. If a check flagged the story, it was edited to fix the problems found, and a separate re-check then reviewed the whole story again.
- Who
- The checks are made by our newsroom, as steps kept separate from the writing, under rules set by our editor, Hussein Mukhtar. A story the checks still flag is held for the editor, who decides whether it is fixed, published or dropped.
- If something is wrong
- “Fact-checked” does not mean error-free. If a material error is found after publication, we correct the story and add a note saying what changed. Report an error
Companies in this story








