The Finance Startup Bringing Agentic AI to Wall Street

Unknown Source July 31, 2025 46 min

artificial-intelligence startup investment google

41 Companies

104 Key Quotes

3 Topics

🎯 Summary

Podcast Summary: Model ML – Building the AI Workspace for Financial Services

This episode features an in-depth conversation with Arnie and Chas Anglander, the founders of Model ML, a Y Combinator (W24) company creating an AI workspace tailored for the financial services industry. The founders are notable for having successfully sold two previous YC companies, Fat Llama and Fancy, before launching Model ML.

1. Main Narrative Arc and Key Discussion Points

The discussion centers on the rapid adoption and tangible value Model ML is delivering to top-tier financial institutions. The narrative moves from introducing the product’s core concept—an agentic “office suite” for finance—to celebrating their recent explosive growth (signing as many contracts in the last seven days as they did in all of Q4). The founders contrast the “testing” phase of AI adoption from last year with the current “using” phase, emphasizing that the underlying LLM improvements (like better function calling and vision capabilities) are driving this shift. A significant portion is dedicated to the founders’ entrepreneurial journey, drawing key lessons on perseverance and hiring from their previous ventures.

2. Major Topics and Subject Areas Covered

Model ML Product: An AI workspace mimicking Word, PowerPoint, and Excel, but built on an agentic system connected to a firm’s internal data (files, emails, CRM, custom datasets) and external sources.
Target Market & Adoption: Financial services, including private equity, investment banks, asset managers, and hedge funds (claiming usage by ~10% of the largest private equity firms).
Value Proposition: Automating highly manual, repetitive tasks like generating earnings summaries from public filings, saving analysts days of work and potentially increasing accuracy by cross-referencing multiple data sources.
AI Model Evolution: The impact of recent model improvements (function calling, vision models/OCR) making agentic systems significantly more capable than they were even a year prior.
Sales Strategy in Finance: The necessity of selling at the CEO/Executive level due to the firm-wide importance of this technology, requiring significant in-person relationship building and trust.
Entrepreneurial Learnings: Insights from Fat Llama (marketplace requiring perseverance) and Fancy (vertically integrated grocery delivery that found instant PMF during COVID).

3. Technical Concepts, Methodologies, or Frameworks Discussed

Agentic System/Cognitive Architecture: The core technical framework of Model ML, designed to mimic human digital access and decision-making processes.
Function Calling: Highlighted as a key technical improvement in newer LLMs that enhances the reliability of agentic workflows.
Vision Models & OCR: Crucial for advanced document analysis, allowing the system to accurately read and interpret data from tables and charts within filings.
Vertically Integrated Model (Fancy): Contrasted with asset-light models, where the company owned inventory and infrastructure (warehouses).

4. Business Implications and Strategic Insights

Shift in Enterprise AI Sales: The financial sector has moved from cautious AI piloting (last year) to immediate, multi-year contract adoption this year, signaling a fundamental change in enterprise software purchasing behavior for high-impact AI solutions.
C-Suite Buy-in: For transformative tools like Model ML, sales cycles require direct buy-in from the most senior executives, not just team leads.
Automated Accuracy: The founders argue that for specific data-gathering tasks (like summarizing filings), LLMs are already proving more accurate than humans who manually trawl documents.

5. Key Personalities Mentioned

Arnie and Chas Anglander: Founders of Model ML, previously founded YC companies Fat Llama and Fancy.
GoPuff: The company that acquired Fancy.

6. Predictions, Trends, or Future-Looking Statements

The tangible value being driven by AI products is accelerating, and this trend is expected to continue quickly.
The product itself is theoretically improving month-over-month simply due to underlying model advancements, even without direct intervention.
The industry misconception is that AI is where it was six months ago; the reality is that lower-level data gathering and presentation tasks are already being fully automated in top firms.

7. Practical Applications and Real-World Examples

Earnings Summaries: Analysts previously spent days compiling a single-slide earnings summary by manually pulling data from filings (source) and external vendors (like FactSet). Model ML now generates a 90-95% complete, sourced summary instantly upon a release.
Investor Due Diligence (Origin Story): The initial prototype automatically generated a one-pager on an investment opportunity by pulling data from email, LinkedIn, CrunchBase, and public filings—information the founders would have gathered manually.

8. Controversies, Challenges, or Problems Highlighted

Trust and Risk in Finance: Selling to senior finance executives involves high personal risk for the decision-maker (“if they make a wrong call here, they could get fired”), necessitating extensive trust-building, often through in-person meetings and detailed demos.
Hiring Difficulty: The founders stressed the difficulty of hiring, emphasizing that cultural fit and enjoyment of the work are now prioritized over CV pedigree.
Fat Llama’s Early Struggle: The initial launch of Fat Llama involved a terrifying incident where a high-value drone rental was nearly stolen, testing the founders’ belief in the core concept and insurance model.

9. Solutions, Recommendations, or Actionable Advice Provided

🏢 Companies Mentioned

PayPal ✅ finance/tech

Michael Seibel ✅ unknown

Paul Graham ✅ unknown

But I ✅ unknown

With Fancy ✅ unknown

North London ✅ unknown

And COVID ✅ unknown

If I ✅ unknown

When I ✅ unknown

New York ✅ unknown

Hong Kong ✅ unknown

A CEO ✅ unknown

Google Drive ✅ unknown

And I ✅ unknown

So Model ML ✅ unknown

💬 Key Insights

"I think building a plane going down the runway is the best way to describe things. That is the quickest way to learn."

Impact Score: 10

"The very worst case—the very worst case—is you're going to learn so much over such a short period of time. And that is the worst case."

Impact Score: 10

""Well, if you're 26 and you're poor, that'll be the worst outcome," and probably most 26-year-olds don't have any money anyways."

Impact Score: 10

"You might be building a business for the next 5, 10, 15 years; it might not go anywhere, and you've got to be okay with that reality."

Impact Score: 10

"Building a business is really, really, really hard. It's really fun if you enjoy that stuff, but it's really hard. So really, you've got to say to yourself, 'Look, you might be building a business for the next 5, 10, 15 years; it might not go anywhere,' and you've got to be okay with that reality."

Impact Score: 10

"There can't be any miscommunication, lack of communication. We all know one of the biggest reasons, if not the biggest reason, why startups fail is founder fallout."

Impact Score: 10

📊 Topics

#artificialintelligence 64 #startup 32 #investment 5