From CBS:
Frank Bisignano, commissioner of the Social Security Administration, told CBS News that he believes technology, and specifically artificial intelligence, could be the key to improving his agency's customer service, despite recent changes that have prompted concern among some of the nearly 69 million Americans that receive Social Security each month.
"We're bringing a massive technology effort to transform the servicing agenda," Bisignano said. "We're gonna bring AI into the phone system...I intend it to be completed this year." …
From the New York Times:
… “The technology we’re building today is not sufficient to get [to Artificial General Intelligence or A.G.I. which would be needed to switch Social Security phone service mostly to A.I.]”, said Nick Frosst, a founder of the A.I. start-up Cohere who previously worked as a researcher at Google and studied under the most revered A.I. researcher of the last 50 years. “What we are building now are things that take in words and predict the next most likely word, or they take in pixels and predict the next most likely pixel. That’s very different from what you and I do.”
In a recent survey of the Association for the Advancement of Artificial Intelligence, a 40-year-old academic society that includes some of the most respected researchers in the field, more than three-quarters of respondents said the methods used to build today’s technology were unlikely to lead to A.G.I. …
The AI phone system that Social Security has tried to implement has been a complete failure so far. I’m not aware of any institution that has an AI phone system that would come close to what Social Security needs. There is good reason to believe that no such system will be in the offing any time soon, if ever. My question is, what’s your backup plan, Mr. Bisignano?
53 comments:
While the idea of using Artificial General Intelligence (AGI) to improve Social Security’s customer service is ambitious, it might not be the most practical solution for the task at hand. As Nick Frosst notes, today’s AI technology is far from the kind of general intelligence we would need for such a broad application. In contrast, narrow AI, which focuses on specific tasks, is already highly effective at handling customer service functions like answering policy questions. AI systems that specialize in these tasks could deliver accurate, timely responses without the need for AGI, which is still years, if not decades, away from being a viable option. Rather than investing heavily in AGI, it would be more efficient and cost-effective to focus on improving and expanding the capabilities of existing narrow AI systems. This could help the Social Security Administration enhance service now, without waiting for a technology that may never fully meet the needs of such a specific context.
(Credit to ChatGPT for this response.)
You don’t need to implement artificial general intelligence to help tremendously with SSA’s phone service in a relatively short period of time. You just need to be able to identify what the caller’s issue is. Being able to triage that work and free up a percentage of our workers’ time is a good thing, not something to rail against. Humans will handle the complex stuff, AI can handle the simple stuff, at least in so far as directing the call to the appropriate staff is concerned. Of course it has to be implemented intelligently, bit by bit rather than swinging for the fences on the first pitch. And there will be glitches, but I don’t know what you call waiting an hour or more to talk to a human or having to call back five weeks later for an actual interview - isn’t that the equivalent of a glitch, or worse? This is the most obvious thing to do for any large organization. The idea that the attempt to do it is problematic is ludicrous.
"Backup plans are plans to fail" or some macho corporate mumbo jumbo is what he'd probably say.
Saying i need to speak to a supervisor to find out what happened to a payment delayed over six months got returned to the main menu, over and over and over. Only by screaming AGENT AGENT AGENT did it get moved to the general hold which lasted over 30 minutes before someone picked up. Then, i was give the new extension number, which was wrong.
Yeah, working great so far!
Charles, there is no backup plan. The AI experiment should have been rolled out slowly but thanks to the DOGE it’s full speed ahead and SSA recipient’s suffer.
Frank’s just another ignorant old man throwing the taxpayers’ money away on magic beans. It’s been the GOP’s favorite pastime since 2001, when they decided technology could save us from terrorism and solve all of society’s other ills if we’ll just piss enough money away on the tech bros.
A timely question, Charles. Tune in this afternoon when Frank gives an address that every single SSA employee will excitedly listen to. We expect him to share his vision.
You’re right. Charles is wrong to rail against Frank because it would actually be great if Frank did the thing you suggested, which is not at all what Frank has proposed.
You Trumpers have olympic level mental gymnastics skills.
It’s a massive failure in leadership, judgment and stewardship of public funds for SSA to use AI to directly interact with the public when it’s proven unreliable for that purpose. The AI bot frequently misdirects, misinforms and frustrates callers. If the leadership doubles down on that failure, then it will be clear that they have little interest in serving the public.
Ask Frank Bisignano, whose net worth is over 800 million, to explain why he took on this demanding job.
Problem is, it fails miserably at directing callers to the correct information or person and the “simple stuff” as you call it. Which they would have known if they cared enough about customer service to test it before deciding to pay for and implement it. Or cared enough to listen to disability advocates and the public who are still pointing that out.
Absolutely correct. AI is a terrible substitute for hiring and training people to handle phone service.
Backup plan is to sit back and enjoy all his profits
The agency already released its own chatbot ASC and SSA employees are raving about it. Many use it daily. So maybe Frank knows what he’s doing?
@10:06. Not a Trumper, dude, far from it. Just tired of doing things in the least efficient way possible and whining all the time. Tired of every single issue, no matter how inherently apolitical, being polarized where you have to choose between Good and Evil. Sometimes it’s just a matter of smart or not smart.
"We're bringing a massive technology effort to transform the servicing agenda.." This is words on a page, with zero meaning. Someone needs to brief Frank on the state of the IT budget and how the lion's share of the agency's budget goes to maintaining Systems' payroll.
Not true 10:42. SSA must provide around the clock service. That can only be done with AI. We must develop outstanding state of the art service tools to alleviate the burden.
Sufficient for summarizing 10 page documents with simple language and structure. Beyond that, it's a freshman-level code generator. No PII of course, so it's useless for complex claims, appeals, manual workload items - the stuff piling up in the backlogs.
This agency is going to get itself sued to oblivion if it forces a hallucinating "AI" program on the public. Congress, stop this madness!
@11:04 ASC may be good for tech help but is terrible for SSA programmatic. The only raving I have heard was needing help with how to do something on a Microsoft product or help write a retirement letter for an employee. Anything asked of actual SSA business has been wrong.
Yeah? I don’t know anyone who’s had a positive experience with it. But cool story, bro!
The SSA chatbot is amazing!! As an OHO decision writer, I can feed it a copy of a decision, after taking the time to redact every bit of PII, and it will spell out the first use of every acronym *and* remove the second space after every sentence! I'm so grateful for that help!
Only a non-SSA would say AI is used by SSA employees and it works. A complete fabrication and outright lie.
Uh, the chat bot is borderline useless. It also raises huge privacy and disclosure concerns. Finally, this bot was released before Frank was confirmed. Are you saying he was approving and directing SSA activity before confirmation?
Anyone seen the Monorail episode of the Simpson's?
@10:35 So what exactly is your solution? Hire more people? Not an option. Also, do you know how many times people get cut off or hung up on or misdirected by actual human beings working for SSA? Or given wrong information? Was everything a well-oiled machine prior to software arriving? Intelligent software is the reality for the future. There is no other serious alternative. Get on board and try to help make it work.
For the tax break once he sells his shares and to give his billionaire buddies contracts to swindle the American public more. I wonder who will be the first to tape this glorious event and leak it to the press
That chatbot pre-dates Frank or Leland, and nobody on the front lines is using it. It doesn’t belong policy, current SSI or T2 rates like SGA or the AET, can’t do household composition or LA determinations or interactive comps. It’s no good to us with claims or PE.
No, they don't. There was a post about it when it was released and people, rightly, laughed at how much of a joke it is.
I'm surprised but glad to hear that the new ASC chat bot is helpful to some. When I tested it out with some policy questions, it gave outdated info on many issues and straight-up made up answers on some.
Can you give examples of what it does well?
Propaganda. The chatbot has nothing to do with Frank, who inherited it but didn't pitch it, design it, or test it. The bot is only programmed with law up to 2023, which makes it useless in terms of instructing the user on disability and overpayment policy, which changed substantially in 2024.
Wrong. The chatbot ASC actually sucks and can’t even get basic SSA information correct. It might be good for composing emails but if you need AI to help you create standard emails you should probably find another job.
Those failures in AI are a feature, not a bug. The end result being that SSA will be contracted out in its entirety. Lets hope it doesn't happen.
Agree with 11:21. Every system “enhancement” makes the job less efficient. It just keeps getting harder and harder to do the simplest things.
I concur
The employees I know ignore the chatbot because it is a worthless joke.
The chat bot is a waste of time. If someone doesn’t know how to use POMS, HALLEX or SHOP, they should find employment elsewhere. Everything is right there, if one can read.
Caller: I have a question about my check.
SSA AI phone system: Understood. You need translation because you're Czech.
Caller: No. My payment hasn't arrived.
AI: You're reporting that you died.
Caller: No! My benefits are late!
AI: Understood. You want your benefits to terminate. Done.
Who uses that turd of a program daily?
Surprising you say that. Both my partner and I used the chatbot and got into arguments with it. I had to correct it with a law. It also failed to tell me what a ruling from 2011 was so important about it that changed a lot of things. I say this who loves using chatgpt out of work
The problem is our customers dont know how to describe what they want. They may think its a missing payment issue but dowm the rabbithole you find out its something else
And some people keep padding their pockets on the next miracle software package the agency purchases.
Frank is a moron, and proved to all of us today that he doesn’t even understand 1% of what SSA actually does. Hard to understand a complex bureaucratic program when you’re too lazy to do any preparation beyond asking google AI for a summary of your job description though.
If someone is seriously using the terrible chatbot to do their job they should be fired for not knowing how to do their job. It's beyond worthless.
11:04 is a MAGA troll. Nobody uses ASC. It’s stupid.
I really hope some good can be accomplished by automating and using AI to make some tasks easier. And I’m all for flattening out the rigid, inefficient hierarchical bureaucracy. But in the end, when machines fail us, it’s going to be people who pick up the pieces and catch the stuff that falls through the cracks. Those who malign and insult the COSS without giving him a chance show how vicious and political his opposition will be. I hope this COSS learns that running what is, at its core, a human service agency as a business has its inherent limitations. Businesses exist for profit to shareholders. I personally would love it if my SSA was as nimble, informative and user-friendly as other websites I regularly use for banking, et al. However, SSA serves all Americans, including the neediest, least sophisticated, most vulnerable, sickest, unhoused, unrepresented and downtrodden who don’t have mobile devices or easy access to computers. We don’t have profits or budget certainty for long-term planning and must follow laws and policy that require due process, fairness, impartiality and good faith in almost everything we do. We also know that performance metrics based on quantity can work against quality and reward managers & employees who know how to “work the stats”. If Bisignano can make any of this better, his time as a public servant will be added to his list of accomplishments.
Reality check for SSA employees…
Anthropic CEO Dario Amodei is confident AI will be a bloodbath for white-collar jobs, and warns that society is not acknowledging this reality.
AI could wipe out up to 50% of all entry-level jobs while spiking unemployment to 10-20% in as little as one to five years, he says. Unemployment is 4.2% in the US as of April 2025.
"We, as the producers of this technology, have a duty and an obligation to be honest about what is coming," Amodei tells Axios. "I don't think this is on people's radar."
From a fired DOGE employee
“I would say the culture shock is mostly a lot of meetings, not a lot of decisions,” Lavingia told Fast Company in the piece, which also noted that he noticed the number of mission-driven people working in government. “But honestly, it’s kind of fine—because the government works. It’s not as inefficient as I was expecting, to be honest. I was hoping for more easy wins.
Exactly. AI is great if the caller can clearly and specifically identify the issue using precise language. Most folks don’t even know what their question is!
Frank is here for himself. He is an narcissistic megalomaniac and proud of it.
@11:28 Would you agree, that only an idiot would buy a car and start driving it if he knew the brakes only worked 50% of the time? There are numerous reports that the AI bot frequently misdirects or misinforms callers with important questions. The mistakes caused by the error-prone AI may not cause physical injuries like a car with bad brakes, but it can cause fiscal injuries when it misdirects and misinforms the public. To add insult to injury, the same public that is being harmed, is continuing to fund this ongoing failed experiment.
You know, a town with money's a little bit like a mule with a spinning wheel. No one knows how he got it, and darned if he knows how to use it.
These people don't know what we do and they don't know what they're doing.
I sincerely hope COSS does well. This agency has been teetering on the brink of ruin for sometime, given the increased claims and dwindling staffing levels. The technology absolutely needs improvement across the agency. COSS has a lot of work to do but has already torched his credibility at SSA. COSS comes off as a guy who is typical of this administration -- arrogant and incurious. They don't know how things work and don't care that they don't know. Never any specifics, only vague corporate jargon about "winning."
Bleak times.
The reorg did not flatten the org structure. It just made it grow vertically by combining components adding another layer. Funny thing regardless components are not stand alone units but still need to work with each other.
Post a Comment