• AI Street
  • Posts
  • UBS Turns Analysts Into Avatars

UBS Turns Analysts Into Avatars

Author

Matt Robinson
May 22, 2025

Hey, it’s Matt. This Week on AI Street:

💰 AI is making “useless” data valuable

🧬 UBS Clones Analysts for Research Videos

📰 WorldQuant CEO on the impact of AI

Forwarded this? Subscribe here. Join readers from the Fed, McKinsey, BlackRock & more.

DATA

AI Is Creating a Market for Siloed Data

AI is running out of training data—even the whole Internet isn’t enough.

We’re so short on data that a new category has emerged: synthetic data, which, as it sounds, is data made by models to train other models.

But the open web doesn’t hold quite all of the world’s knowledge.

Data behind paywalls, in proprietary research, old databases, and dusty archives is now more valuable.

The co-head of investment banking at Goldman, Kim Posnett, recently wrote about how this data might be monetized in the future, leading to new data marketplaces as she wrote in The new markets for AI data.

There is a lack of sufficient training data that shows humans “showing their work” in the steps to address complex problems. This is where companies with focused, well-organised, or highly logical data sets can become newly relevant. Imagine how a textbook company might use its archives of technical manuals and coursework to train an AI system to do complex scientific processes.

Recent data licensing deals show how different companies are selling access to their data to AI companies. Expect this trend to accelerate as companies get even more creative in doing so. So far, these deals have been negotiated individually with special terms, but you can imagine a marketplace — or multiple markets — for training data emerging.

Kim Posnett, co-head of investment banking at Goldman Sachs

B2B data is more locked away than B2C and companies are rightly worried about how AI companies will use it. We’re still in very early stages. There are no clear guidelines yet for how regulated data could be bought and sold.

But the idea of AI data marketplaces doesn’t seem so far off.

YouTubers are already selling their data to AI companies and getting nicely compensated for it:

AI companies are currently paying between $1 and $4 per minute of footage, the people said, with prices increasing depending on video quality or format. Videos that are shot in 4K, for example, go for a higher price, as does non-traditional footage like videos captured from drones or using 3D animations. Most footage, such as unused video created for networks like YouTube, Instagram and TikTok, is selling for somewhere between $1 and $2 per minute.

Hundreds of creators are participating, with one licensing company paying out over $5 million. Talent agencies like CAA and Wasserman are brokering these deals, while specialized licensing companies aggregate thousands of hours from multiple creators to sell in bulk.

Not bad for footage that was left on the cutting room floor!

Takeaway

AI is turning “useless” content—from training manuals to TikTok drafts—into a sellable asset. Companies may be sitting on valuable intellectual property without realizing it.

SEND IN THE CLONES

UBS Turns Analysts Into Avatars

Made with Ideogram

I was not expecting this:

UBS Bank is creating videos featuring realistic AI versions of its analysts to share with clients, in addition to the research notes it disseminates.

The Swiss bank started the project at the beginning of this year and wants to ramp up the use of AI to assign its analysts more productive projects.

The video undertaking, done with OpenAI and Synthesia models, already has the buy-in of over 36 analysts (out of 720, per UBS’ website) and plans to roll out the capability globally, UBS told Fortune.

….

Creating a video starts with digesting the analyst’s reports and creating a script. The UBS analyst then signs off on the script before it’s made into a video.

The bank currently produces about 1,000 videos annually due to studio constraints. With avatars, UBS aims to increase this to 5,000 videos per year, meeting the rising client preference for video content.

Takeaway

I'm not sure what's stranger: the fact that we're already so comfortable with avatars or that people want TikTok videos of equity research?

Regardless, this seems like.. it’ll be the norm in a few years?

Editing videos is not fun and doesn’t require a ton of critical thinking. Analysts are freed up to focus on bigger projects or meet with clients in person, according to a UBS spokesperson.

QUANTS

WorldQuant founder Igor Tulchinsky is betting big on LLMs. He says they help uncover new “alphas” by structuring messy, unstructured data. “It’s like a free lunch,” he told Forbes.

I’ve said before that even if AI doesn’t get any better, its ability to organize random documents will have a massive impact on companies that will take years to digest.

Some background on Tulchinsky:

  • Started his career at AT&T Bell Labs, then joined Timber Hill (the precursor to Interactive Brokers) under Thomas Peterffy

  • Later teamed up with Israel Englander at Millennium, before spinning out WorldQuant in 2007

  • Known for statistical arbitrage and managing $23B across WorldQuant and WorldQuant Millennium Advisors

He had some pretty big statements on AI’s impact on the quant world:

“We can be using AI and LLMs to convert and discover alphas in different domains,” he says. “Possibilities are endless. The LLMS are getting stronger and stronger.”

With the proliferation of data and AI, we’re getting to the point where you can kind of quant everything.

One thing I would have liked to hear more about in the story is how do you implement this? How much computing power do you need? How do you protect your internal data? Granted this is a Forbes story, they’re not writing so you can build an LLM-quant shop. Also, nice to see my podcast (Alpha Intelligence) co-host Francesco Fabozzi quoted in the article!

Takeaway

These are pretty bold quotes from Tulchinsky. I think initially the AI boom has centered on equity research tools for fundamental analysts (that’s where I started), but I’m beginning to think that AI will have as great an impact on quant land.

This reminds me of what, Tobias Moskowitz AQR Principal, told Bloomberg a couple weeks ago:

“I’ve always been a big skeptic on being able to time factors and other things,” he said, referring to the practice of predicting when each factor will do well — a practice Asness once called a “sin.” “I’m now convinced that we can do a little bit of it.”

Tobias Moskowitz, AQR Principal
HIRING

There were quite a few stories this week about AI slowing hiring. Most of these are anecdotal. I’ve not seen any reliable data that AI is curbing hiring in the aggregate. (If you have, please reply to this email with a link.

  • The AI Hiring Pause Is Officially Here (Bloomberg)

  • JPMorgan Chase wants bosses to 'resist' hiring and use AI (Yahoo)

  • How AI Has Already Changed My Job (Bloomberg)

  • AI may shoulder up to 40% of workload (Banking Dive)

The Fed Asked How GenAI Is Being Used—Here’s What They Heard

The San Francisco Fed recently convened a roundtable with executives from companies in consumer goods, apparel, agriculture, and food distribution to discuss how generative AI is being deployed inside their businesses. (Fed)

  • Speed gains are real. One executive said their company is now operating twice as fast as it did two years ago.

  • No (current) layoffs. Most execs described GenAI as a way to augment employee output, not reduce headcount.

  • Back office impact is already showing. Accounting and IT teams are reporting measurable productivity gains.

ICYMI: MARKETS EDITION

In the last Sunday’s AI Street Markets, I wrote about a free tool that connects Claude to Yahoo Finance—no API key required.

Thanks to Model Context Protocol (MCP), AI can pull the structured data, like stock prices, without custom code.

MCP acts as a bridge between probabilistic LLMs and traditional code.

Investor-agent is an open-source example built on MCP. You can ask Claude things like “What are Tesla’s key financials?” and get a live dashboard back—with revenue, margins, insider activity, ownership data, and technical indicators.

It uses real-time data from Yahoo Finance.

The tool has limitations. It struggles with big calculations (like scanning the entire S&P 500), and Yahoo data occasionally cuts out.

The tool is free, easy to clone from GitHub, and doesn’t require you to write code—especially if you use AI dev assistants like Cursor or Windsurf.

If you missed the walkthrough, it’s worth a look. These kinds of tools are getting better fast.


Takeaway: Free tools are getting better for stock analysis.

DIRECTORY

AI Tool Directory

I’ve put together a database of 50+ AI and investing tools categorized by use case, where you can search, for example, by due diligence or AI agents or earnings data. Each company card has a short description and a link to the platform.

I did my best to add all relevant companies, but I know there are probably more out there. If you know of any that should be included, please reach out: [email protected].

I plan to add more data on each company, but I wanted to see what interests readers most first. So please reach out with comments!

To access this (free) database, please recommend AI Street to someone you think will get value from reading AI Street. (Please don’t send this to a burner email; I can see where the referrals go 🙂)

Share the link below and after they join, you’ll receive the database by email.

WHAT ELSE I’M READING 

OpenAI to Buy AI Device Startup From Apple Veteran Jony Ive in $6.5 Billion Deal (Bloomberg)

AI is improving more quickly than we realize (Bloomberg)

Brokers Sidestep Market-Making Giants With Private Rooms (Bloomberg)

Apple to Open AI Models to Developers, Betting That It Will Spur New Apps (Bloomberg)

Aveni Launches UK’s First LLM for Financial Services (Lloyds)

Citi launches AI tools for Hong Kong employees (Reuters)

Canadian lender RBC sets up new AI team for capital markets unit (Reuters)

How did you like today's newsletter?

Login or Subscribe to participate in polls.

Reply

or to participate.