Wall Street Banks Pay $25,000 a Day for AI Fluency
Bloomberg’s Sally Bakewell argues that Wall Street’s AI challenge has shifted from buying software to teaching bankers how to use it in finance-specific work. She says firms have already spent heavily on AI tools, but demand is rising for trainers such as Wall Street Prompt, which can charge $25,000 a day to teach bankers how to apply generative AI to tasks such as founder diligence, earnings analysis and forecasting. In Bakewell’s account, banks are treating AI fluency as a competitive necessity as much as a productivity initiative.
Wall Street’s AI bottleneck is fluency, not procurement
Wall Street’s AI problem, as Sally Bakewell framed it, has moved past procurement. Banks can buy the tools, and Bakewell said they have already invested “millions and billions” in AI software. The harder problem is whether senior professionals know how to use those tools well enough to remain productive, and whether that fluency helps keep the institution competitive.
Bakewell described the moment as an “AI reality check” for Wall Street: AI is no longer being treated only as an efficiency tool, but as “a requirement for survival.” That view sits inside a broader tension Caroline Hyde raised at the outset: banks may be apologizing for calling certain people “lower value human capital,” but they are still committing to training. In other words, automation anxiety and training spending are not opposites. They are becoming part of the same institutional response.
AI is not just a tool for efficiency anymore. It is also a requirement for survival.
That shift is creating a market for trainers who can teach bankers not simply to experiment with generative AI, but to apply it to the work banks already do. The clearest example was Wall Street Prompt, a firm set up by two former SoftBank fund managers. Its proposition, in Bakewell’s telling, is direct: teach elite bankers how to stop their jobs from being automated.
The price is not the only signal of demand. Bakewell said the firm has a two-month backlog. For banks, the willingness to pay that rate suggests the perceived scarcity is not the technology itself, but people who can translate AI tools into finance-specific workflows.
Wall Street Prompt is selling workflow fluency, not generic AI literacy
Ed Ludlow pointed out that Wall Street Prompt is less than a year old and already has “real customers.” Bakewell said the firm was founded last year and has clients including Bank of America, Citi, and T. Rowe Price.
Its training sessions, she said, typically involve 25 to 30 employees inside banks. The examples she gave were not abstract lessons in prompting. They were applied to specific banking tasks: using Google Gemini together with FBI-style behavioral analysis to spot red flags in founder pitch videos, and using ChatGPT and Claude to analyze earnings transcripts for the most market-moving information and build financial forecasting models from that analysis.
That distinction matters. The training being described is not basic tool orientation for people who have never opened a chatbot. It is instruction in how to use generative AI against recognizable finance materials: pitch videos, earnings transcripts, market-moving information, and forecasting models.
The commercial point is clear: the value being sold is not access to Gemini, ChatGPT, or Claude. It is the ability to use those systems in ways that map onto the tasks Bakewell named. Wall Street Prompt’s pitch depends on the possibility that, if bankers do not learn those workflows themselves, the workflows may be automated around them.
Singapore is being treated as a competitive signal
Bakewell said Wall Street Prompt is considering expansion in Singapore because of the city-state’s focus on ensuring that people entering the financial sector are “very AI fluent.” She described that as part of a competitive edge in Asia, with Singapore in particular putting pressure on US financial firms to raise their own standards.
The Singapore angle turns AI training from an internal productivity initiative into a competitiveness issue. If financial centers differ in how quickly their professionals become fluent with generative AI, banks may face pressure not just to adopt tools, but to make AI capability a common part of finance-sector labor.
That concern fits the broader behavior Bakewell described across large financial institutions. JPMorgan has rolled out an “LLM suite,” which she described as a generative AI tool. Goldman is working with Anthropic. Bank of America says AI has made its developers more productive. Jamie Dimon, Bakewell added, says he uses AI every day.
Those examples point to a sector that already sees the need to upskill staff. The anxiety is not that AI has arrived outside finance and banks have failed to notice. It is that investment in tools and partnerships does not automatically create AI-capable professionals across the organization.
Automation anxiety is becoming a training budget
Wall Street Prompt’s pitch gives concrete form to a larger institutional fear: AI may change which banking tasks require human labor, but banks still need people who can use the systems well enough to preserve productivity and judgment inside the firm. Bakewell’s account suggests that finance firms are increasingly treating AI fluency as a professional requirement rather than a technical specialty limited to developers or data teams.
The training market she described is built around finance-specific instruction. Bankers are being taught to use AI to spot red flags in founder pitch videos, extract market-moving information from earnings transcripts, and build forecasting models from that analysis. Those are not peripheral uses. They sit close to the work of diligence, research, and financial judgment.
Banks can possess sophisticated AI systems and still face a practical gap if senior professionals do not know how to use them in the workflows Bakewell described. In her framing, the competitive question is not only which institution has bought the tools. It is which institution has enough professionals who can put those tools to work before the work is automated around them.