A top executive at a major Wall Street bank is deep into his spiel on how artificial intelligence (AI) will make the firm smarter and leaner when he pauses to take a question: What does this mean for young people entering the business?
The silence grows."It's, um," he says, shifting tone and making clear he can't speak publicly. That question is gnawing on him, he confides, because he has kids. "I would want them to pick their careers very carefully. I think AI is going to eliminate most jobs. That's a private view. I think we're just starting to feel that."
Within the upper echelons of many financial firms, there's a lot of soul searching as executives prepare to roll out a new generation of technology. Publicly, they are upbeat, predicting machines will perform almost all repetitive tasks, freeing humans to focus on more valuable pursuits.
Privately, many confide to peers, consultants and sometimes journalists that they are worried about what will happen to their staff - and what to tell them.
There's also uncertainty. Maybe it's all overblown, executives say, because the tech will be hard to implement and humans will find new roles. Or perhaps it's the beginning of the end for legions of professionals in one of the world's most lucrative fields. Can jobs held by office-dwelling millionaires disappear like those on factory floors?
The result is that employees aren't getting a clear message on what's to come.
For a rosy scenario, look to McKinsey & Co.
Early adopters like Goldman Sachs are working with tech, including machine-learning software that improves itself by searching data for patterns.
In July, the consulting firm published a report estimating machines are ready to assume roughly a third of the work now performed by banks' rank and file. The authors framed it as positive: People will have more time to tend to clients, conduct research or brainstorm ideas. So far, it noted, firms at the forefront aren't slashing jobs.
At JPMorgan Chase, one of the most tech-savvy banks, chief executive officer Jamie Dimon predicted in June that his workforce will more likely grow than shrink over the next 20 years. Technology may displace workers, he has said, but it also creates opportunities.
Yet in interviews, about a dozen Wall Street executives and consultants responsible for deploying technologies - and steeped in their capabilities - were more bearish on humans. Machines will take over task after task, they said, and banks simply won't need nearly as many people.
It's time for senior managers to stop sugarcoating, said Mr Simon Moss, who has been advising banks and investing firms as head of Grant Thornton's fintech and innovation practice for the industry."Are there positions in financial services that are actually untouchable from technology? The simple answer is 'no'," Mr Moss said. "It's just a case of when."
Early adopters like JPMorgan, Goldman Sachs and Bank of New York Mellon Corp are experimenting, he said. They are working with tech, including machine-learning software that improves itself by searching data for patterns, and natural language processing, which helps computers comprehend human speech.
Once firms figure out how to deploy those, rivals may quickly follow suit. Humans will have less to do, and banks will whittle costs by shrinking headcount."Sitting with excess capacity is not a good business decision," Mr Moss said. Managers should start warning employees that they need to learn more skills to stay on.
That's supported by history. Floor traders who once shouted orders in raucous marble-columned exchanges have seen their numbers dwindle since the 1990s, forcing many to seek new work.
Mr Gary Cohn, who last year stepped down as president of Goldman Sachs, told investors in 2011 that technology had helped the firm slash equities staff by more than half over the previous decade. The trend was similar in currencies.
Few automators set out to kill jobs.
Mr Inder Thukral, co-founder of Kognetics, is making machine-learning algorithms that ease life for investment bankers working around the clock on mergers and acquisitions.
"It's the spirit-crushing work that it's replacing," he said in an interview. Dealmakers should focus on only the most important stuff, he said. "Searching through 6,000 documents is a computer's job, so let it do it."
Ultimately, that means fewer people will be needed, he said. But if planned well, banks may be able to avoid firings through attrition - leaving positions open as they become vacant.
That hope comes up a lot.
At UBS Group, executives are using a digital employee known as Amelia to quickly perform repetitive tasks in the firm's investment bank and asset management divisions. Executives say employees shouldn't fret as the software, developed by digital labour company IPsoft, spreads across operations.
"The first question at mind is 'What happens to my job?'," said Mr Tom DeCarlo, a managing director who sits on the bank's innovation committee. But dismissals aren't his plan. "As we have attrition - internal or external - I'm just not replacing those people."
Mr Sergio Ermotti, the bank's CEO, said in a recent interview that the firm could shed 30 per cent of its employees over a decade.
Another senior UBS executive, asking not to be quoted by name, estimated that technology would allow it to operate with up to 40 per cent fewer people in just four to eight years.
Attrition has limits. Some employees may retire, while others quit to take more promising jobs elsewhere. But if the whole industry is shrinking, those opportunities may be scarce.
There's no denying that fewer people will be needed, said Mr Richard Johnson, vice-president of market structure and technology at Greenwich Associates. At least, the transformation will be incremental. "It's not as if everyone is going to come to work and there's a bunch of robots sitting in their seats," he said.
But eventually, a transformation is coming, said IPsoft CEO Chetan Dube, who is working with UBS and more than 100 other finance firms to deploy his company's software. "Time, tide and technology will wait for no one," he said. "We've tippy-toed around this topic too much and for too long."