ROBOTS

AI Agents Arrive at Citi

“Agentic” roughly means giving AI autonomy to carry out a task you define without ongoing supervision. The agent can monitor for new information, decide what’s relevant, summarize or analyze it, and trigger follow-up actions.

For example, you could set up an agent to track JPMorgan’s 8-K filings, extract any updates on credit loss provisions, compare them with previous quarters, and send you a short analysis of what changed.

On the consumer side, you could ask an agent to: ‘Book me a nonstop roundtrip flight from Milan to New York, departing Nov. 15, for one week.’ (Google, btw, is developing an agentic payment protocol to make this happen.)

Corporate adoption of AI agents is just beginning.

From the WSJ:

Citi is one of the few major companies, alongside Walmart and Microsoft, putting AI agents into live tests.

Citi’s internal platform, known as Citi Stylus Workspaces, uses models including Google’s Gemini and Anthropic’s Claude. It lets users research a client, compile a profile from public and internal data, and translate it into another language in one step.

Citi Chief Technology Officer David Griffiths told the WSJ that the bank has built cost limits into its systems, but falling model prices make ROI hard to calculate and longer-running tasks could eventually change the cost equation.

Relying on the biggest models can get expensive pretty quickly and is likely not necessary for a straightforward task.

I reached out to Citi for more detail on whether they’re using smaller models, but they passed on commenting.

Takeaway

The current AI boom is often seen through the lens of chatbots that can write you a sonnet or draw a cartoon. Now, the technology is slipping into the corporate background and becoming part of the plumbing of white collar work.

Related

Amazon’s Finance Teams Are Relying More on AI—and Not Just for the Simple Stuff (WSJ)

NYC

AI Street Meetup

Thanks to all those who were able to make it to the first AI Street Meetup. It was great to meet many of you in person! And thanks to my co-host Jordan Hauer for his help in arranging.

CONFERENCE ROUNDUP

AI in Finance: Takeaways from Cornell’s Future of Finance Conference

I attended Cornell’s Future of Finance Conference in New York City last week. The event brought together industry and academic voices from firms including Citadel, Millennium, Balyasny, IBM, and XTX to discuss how investors are adopting AI, quant modeling, and alternative data.

It’s always great to hear from folks actually using a technology as hyped as AI is. While hedge funds understandably keep in-house developments private, most of the conversation stayed at a fairly high level. My sense is that we’re still in the early stages of AI’s evolution on Wall Street.

But rather than just sharing my takeaways, I asked some attendees (who also joined the AI Street meetup I hosted) what stood out to them in the current state of AI in finance.

Dan Averbukh, CEO of validityBase

• My biggest takeaway is that the buy side is extremely suspicious that the models that are being built with AI can generate real alpha. If you want to succeed with building AI for buy side participants, you’ll need to bring receipts.

William Wu, CEO and CoFounder of Menos AI

• The power of GenAI in investing is now consensus. But moving from promising prototypes to production is proving far harder than expected and that’s the transition where firms will either build lasting edge or fall behind.

Rodger Coyne, Founder of VecViz

• Comments suggested that many historically discretionary investors have gotten themselves somewhat "tooled-up" with models and data and are now working out how the resulting analytics fit in their investment process.

Sarah McKenna, CEO of Sequentum

• It seems AI is still very early and a very blue ocean!… By September 2025 of "The Year of AI Agents" I was expecting some pretty illuminating anecdotes about how AI is reshaping the industry. Instead I came away struck with how much everyone is still experimenting.

• I did hear one very senior technologist say his hedge fund had maxed out [their data cloud platform] ability to scale and they had moved away from [that provider] as a consequence and I thought that reflected the scale at which they are experimenting (with no magic bullets) but also how big a threat AI is to the incumbents. It does sound like finance teams have had success in AI agent workflows like classifications but nothing yet at scale.

Sangmin Seo, Quant Researcher at Arrowpoint Investment Partners

• There was no doubt that AI in finance is a new wave with a lot of opportunities. At the same time, everyone warned that just "using AI" is not the way to go. Who knows how to use AI wisely and safely will dominate the market.

SPONSORSHIPS

Reach Wall Street’s AI Decision-Makers

Advertise on AI Street to reach a highly engaged audience of decision-makers at firms including JPMorgan, Citadel, BlackRock, Skadden, McKinsey, and more. Sponsorships are reserved for companies in AI, markets, and finance. Email me ([email protected]) for more details.

ADVISORY

ChatGPT, What Stocks Should I Buy?

Reuters reports that a growing number of retail investors are relying on ChatGPT for stock picks.

The idea of AI investing for you might seem far-fetched to anyone familiar with the “thundering herd,” but for younger investors, AI feels natural.

About half of retail investors say they would use AI tools such as ChatGPT or Google’s Gemini to pick or adjust investments, and 13% already do, according to an eToro survey of 11,000 retail investors worldwide. That figure appears to come from July 2023, so actual adoption is likely higher today.

You can already invest in ChatGPT investment ideas.

You can already invest in ChatGPT investment ideas. On Autopilot, the GPT Portfolio is a model portfolio that uses ChatGPT to select 10 stocks and 5 sector ETFs and rebalances monthly. Autopilot charges a flat subscription fee rather than a percentage of assets. The fund is up 17% so far this year, compared to the S&P 500’s 13% gain.

Takeaway:

AI money managers are likely to move from novelty to reality as younger investors, already used to buying their beds and cars from their phones, start trusting them with portfolios.

ADOPTION

Fed Data Points to Slower Hiring in AI-Exposed Jobs

Two recent Fed studies show AI’s labor impact is showing up first in slower hiring, not layoffs.

The St. Louis Fed found that from 2022–2025, occupations most exposed to AI, tasks that could be automated and jobs where workers use AI, saw the largest unemployment increases, especially in computer and mathematical roles. The authors stress this is a correlation, not proof of causation.

The New York Fed said just 1% of service firms reported AI-driven layoffs in the past six months. A few employers expect AI-driven layoffs soon, but the Fed noted that similar predictions last year didn’t really materialize. Many are retraining workers or hiring for AI skills instead of cutting staff.

AI adoption remains concentrated in information, finance, and professional services. With only 25–40% of firms using AI, researchers say the economy-wide effect is still modest.

UBS chief economist Paul Donovan wrote last week that the spike in U.S. youth unemployment is unique globally and not caused by AI despite public debate over automation. He noted that youth unemployment is falling in Europe, the UK, and Japan, making it unlikely that AI is uniquely harming young U.S. workers.

Takeaway

There’s currently no convincing data that AI is coming for Wall Street jobs. This could be in part because we’re still early in adoption, but I’m skeptical. Evidence points to employers retraining workers rather than cutting staff.

Related

• Anthropic Finds Businesses Are Mainly Using AI to Automate Work (Bloomberg)

WHAT ELSE I’M READING

• The Evolution of AI Agents in Finance (Didier’s Newsletter)

• Meta Pushes Into Power Trading as AI Sends Demand Soaring (BBG)

• An $800 Billion Revenue Shortfall Threatens AI Future: Bain (BBG)

• Nvidia, OpenAI Make $100 Billion Deal to Build Data Centers (BBG)

• BoE’s Bailey: AI Could Help Regulators Spot Hidden Risks (Reuters)

• Big tech's AI push is driving U.S. business investment (MarketWatch)

• VCs to AI Startups: Please Take Our Money (BBG)

• Tilt Raises $7.1M Seed Round For AI-Powered Direct Indexing (PR)

• Trading Venue OneChronos Taps Ex-S&P CEO for Growth (BBG)

• Ally CIO Q&A: Pace of tech change ‘weighs on me’ (Banking Dive)

• FactSet Falls as Wells Fargo Cuts Target on AI competition (Investing)

• Study: AI Passes CFA Level III (Wealth Management)

• HSBC, IBM Trial Quantum-Enabled Algorithmic Bond Trading (PR)

• Modular Financial Agent Systems (II) (Encyclopedia Autonomica)

• D.E. Shaw – the quant king (Rupak Ghose)

CALENDAR

Upcoming AI + Finance Conferences

• Bloomberg-Columbia ML in Finance Conf – Sept 25, 2025 • New York

  Academic–industry event hosted by Columbia University and Bloomberg, focused on ML applications in finance including asset pricing, market forecasting, and LLM risk.

• GAIIM Conference 2025 – Sept 30, 2025 • New York

  Forum on practical applications of AI in investing, featuring tools for research, valuation, and portfolio workflows.

• Open Source in Finance Forum - Oct. 21-22 • New York

  Finance and tech leaders tackle how open source and AI can be governed, scaled, and applied in financial services.

• AIFin Workshop at ECAI 2025 – October 26, 2025 • Bologna, Italy

  One-day academic workshop on AI/ML in finance, covering trading, risk, fraud, NLP, and regulation.

• AI in Finance 2025 – October 27–30, 2025 • Montréal

  Academic event covering ML in empirical asset pricing and risk.

• ACM ICAIF 2025 – November 15–18, 2025 • Singapore

  Top-tier academic/industry conference on AI in finance and trading.

• AI for Finance – November 24–26, 2025 • Paris

  Artefact’s AI for Finance summit, focused on generative AI, future of finance, digital sovereignty, and regulation 

• NeurIPS Workshop: Generative AI in Finance – Dec. 6/7 • San Diego One-day academic workshop at NeurIPS focused on generative AI applications in finance, organized by ML researchers.