On Friday, the Social Security Administration rolled out a new ChatGPT-style chatbot called āAgency Support Companion.āĀ The tool was meant to help staff handle routine tasks and boost productivity. Instead, employees say it barely works, and a mismanaged launch has left many wondering if automation is the right path for Americaās largest social services agency.
In recent months, Muskās organization, known internally as DOGE, has aggressively reduced federal agencies by cutting jobs and budgets. The push toward an āAI-firstā model aims to replace human workers with software wherever possible. As part of that strategy, the new chatbot was built into the everyday workflows of agency staffers, reported by Wired, with the promise of easing their workload.
An internal email described it as a way to āassist employees with everyday tasks and enhance productivity,ā but in practice, the app has received little attention.
Sources inside the agency say the roll-out was accompanied by a painfully bad training video featuring a poorly animated, four-fingered woman. The clip was supposed to guide users through the appās features, but it failed to warn staff against uploading sensitive personal data. This slip forced the agency to send a follow-up email reading, āOur apologies for the oversight in our training video,ā to warn employees of the risk.
Few people have taken the SSA AI training video seriously
āIām not sure most of my coworkers even watched the training video,ā one Social Security staffer told Wired. They added that when they tested the chatbot, āseveral of the responses I received from it were incredibly vague and/or inaccurate.ā Other employees reportedly mocked the crude graphics. āNobody I know is [using it]. Itās so clumsy and bad,ā the source said.
Critics warn that this stumbling start may foreshadow wider failures.
Similarly, in Brazil, during 2018, a state-owned company called Dataprev launched Meu INSS, an app designed to process social security claims using computer vision and natural language processing. However, the system has often rejected valid claims due to minor errors, triggering lengthy legal appeals.
One example involves 55-year-old JosĆ©lia de Brito, a former sugarcane worker whose retirement application was denied because the automated system misidentified her gender. āI have all the documents proving my health condition, proving everything, and [the benefit] still gets denied. Itās a humiliation,ā she told the Rest of World.
Rural farmworkers, many of whom lack basic digital skills, have struggled with the shift to online services. āPeople out here cannot [even] work with Gmail, Facebook, Instagram,ā said Francisco Santana, president of the Union for Rural Workers at Barra do Corda. āProcesses are [getting] more and more automated, and society wasnāt made ready for it, especially further away, in the outskirts, for people that live in rural areas.ā
Those hurdles abroad highlight potential pitfalls for the U.S.
Social services are already under pressure to āmodernize,ā a push that began under the Trump administrationās team of so-called technocrats.
Muskās DOGE has taken that mission further by seeking to cut the federal workforce in half. āEverything that can be machine-automated will be. And the technocrats will replace the bureaucrats,ā a person familiar with DOGEās plans told the Washington Post.
DOGEās emphasis on automation has bred confusion rather than efficiency.
A recent mishap at the SSA illustrates the stakes. DOGE staff reportedly marked ācountlessā living benefit recipients as ādead,ā cutting off their payments. āAbout 4 million people, they marked them as dead,ā said Rennie Glasgow, a longtime claims technical analyst at the agency, speaking to The Daily Beast. āTheyāre sending us an email saying, āIf these people come into the office with their identification, you can reinstate them.āā That reinstatement process can take āabout three to four days,ā Glasgow added, as staff must manually āresurrectā each recipient in the system.
Beyond stopping benefits, DOGE also plans to rewrite the SSAās entire software codebase within months. Sources say that to meet that deadline, the team will almost certainly rely on AI for coding, a practice that demands close oversight because of the high error rate in machine-generated code.
Given DOGEās rocky track record so far, an automated overhaul could introduce even more chaos.
Some observers suspect those mistakes may be intentional. A school of critics argues that DOGEās goal isnāt genuine reform but rather the destabilization of the SSA to pave the way for privatization. If so, the misfires and mayhem may be exactly what Muskās team is aiming for.
#AI #DOGE