--- title: "What Is Model ML? The AI Operating System for Finance Explained" url: "https://goremy.ai/blog/articles/model-ml-ai-operating-system-finance" author: Luis Chavez-Mattos editor: Luis Chavez-Mattos editor_title: Director, Product published: "2026-11-13T13:00:17.082Z" updated: "2026-11-13T19:56:46.740Z" topic: AI Models & Industry description: Model ML builds AI agents for banks, asset managers and consulting firms. Its funding, growth, Excel breakthrough and what it means for finance jobs. --- # What Is Model ML? The AI Operating System for Finance Explained > Model ML builds AI agents for banks, asset managers and consulting firms. Its funding, growth, Excel breakthrough and what it means for finance jobs. Edited by Luis Chavez-Mattos, Director, Product ยท Published 2026-11-13 ## What is Model ML? Model ML is a Y Combinator-backed startup that says it is building "the AI operating system for financial services." It sells AI agents to large banks, asset managers and consulting firms to automate work that analysts do by hand, such as due diligence, M&A support and other investment banking tasks. The company came out of stealth about 18 months ago, has roughly 200 employees, and says it has raised a little over $150 million. ## TL;DR - Model ML is **18 months old** and grew from about 5 to 10 people at launch to roughly 200 today, adding around 20 people a week according to its CEO. - The company says it has raised "there or thereabouts" **over $150 million** and runs eight offices in major financial hubs. - Its customers are the largest banks, asset managers and consulting firms, which usually take **6 to 12 months to buy** new software. - The CEO says AI agent tools **could not work inside Excel until February 2026**, and that this made earlier versions of the category close to useless for finance. - The founders came to the idea after going through **two painful transactions** where they typed out answers to long diligence question lists. - For finance professionals, the practical shift is that routine modeling and document work is becoming something you review, not something you build from scratch. ## Who is behind Model ML? Model ML is the third Y Combinator company for its CEO and his brother Arie, who is the technical co-founder. Their earlier startups were Fat Llama, a marketplace for renting items from people nearby that joined YC in 2017, and Fancy, which joined in 2020. They sold both. The CEO studied accounting and finance and taught himself to code. That mix matters here. The product is aimed at people whose daily tools are spreadsheets, data rooms and slide decks, and the founders have sat on the receiving end of those workflows. The idea did not start as a company. After selling their second business, the brothers ran a small family office, investing their own money and experimenting with large language models (LLMs, the technology behind chatbots like ChatGPT) on their own work. For a couple of years they had no plan to turn it into a product. By the end of 2024 they decided to, having applied to YC at the end of 2023. ## How big is Model ML right now? These are the figures the CEO gave in an interview. They are self-reported and approximate. | Item | What was said | |---|---| | Age | About 18 months since leaving stealth (last February) | | Headcount at launch | Roughly 5 to 10 | | Headcount today | About 200 | | Hiring pace | Around 20 people a week | | Funding | Over $150 million, "there or thereabouts" | | Offices | Eight, in major financial hubs | | Customers | Large banks, asset managers, consulting firms | The CEO also said the company went from 10 to 200 people in nine months, so the "veterans" had been there about four weeks. His lesson from that pace is practical: when you add heads this fast, you must be very clear about where people go for questions and new information. ## Why did Excel support in February 2026 change things? Finance runs on Excel. Valuation models, deal comparisons, budgets and forecasts live in spreadsheets, often with formulas that link across many tabs. An AI tool that can write a summary but cannot open, read and edit a workbook is limited in a bank. The CEO said that agent harnesses (the software that lets an AI model take actions with tools, not just chat) could not interact with Excel until February this year. His words for the period before that: "everything prefebruary was sort of useless." That is a founder's view, and a self-interested one, but the logic holds up. An agent that cannot work in the tool your analysts live in cannot do their job, only advise on it. Once an agent can open a model, trace a formula, change an assumption and produce an updated file, the work it can take on gets much closer to what a junior analyst does. ## What work are the agents automating? The interview excerpt centers on due diligence and "M&A work or investment banking type work." The founders felt the pain themselves across two transactions. A buyer or seller sends a large bank of questions, and someone has to type out answers by hand, pulling from documents and models. The CEO called this "so ridiculous" given what software can now do. Model ML's agents are also described as doing three broader jobs: 1. **Monitoring companies.** Watching a set of companies for news, filings or changes worth acting on. 2. **Building LBO models.** An LBO (a buyout financed mostly with debt) model is a spreadsheet that estimates the returns from buying a company with a large amount of borrowed money. Analysts spend hours building and checking them. 3. **Writing pitch decks.** Drafting the slide presentations that bankers use to win or present deals. One workflow that fits this pattern is monitoring for take-private candidates. A take-private is when a buyer acquires a publicly listed company and delists it. A monitoring agent could keep watching a list of listed companies and flag the ones that start to look like targets, so a human reviews a shortlist instead of scanning everything. The interview material I reviewed does not lay out Model ML's exact steps for this, so treat it as an illustration of the category, not a description of their product. ## Why does the CEO call this "winner takes all"? Selling to big banks is slow. The CEO said procurement at the largest banks, asset managers and consulting firms takes 6 to 12 months. Because Excel-capable agents only became possible in February, he argues the buying decisions are being made now, and the vendors that get through procurement first will be hard to dislodge. That is why he said it is "kind of winner takes all in December" and said December 2026 may look like the most intense stretch in hindsight. It is also why the company hires so fast. The founders described a first year of working 9am to 9pm, seven days a week, and said the AI era has "turned up a notch" compared with their earlier companies. ## Is Model ML replacing investment bankers? Nothing in the interview claims that. The pitch is about automating specific tasks: answering diligence questions, building standard models, drafting documents. The people the CEO described hiring include Harvard MBAs, Stanford and Oxford graduates and people coming out of tier one investment banks, which suggests the company wants finance expertise next to the engineering. Still, the entry-level work in banking and consulting is exactly the repetitive, spreadsheet-heavy work these agents target. That affects how junior roles are structured, even if senior judgment, client relationships and deal negotiation remain human. ## What does this mean for you? **If you work in finance, banking or consulting.** Expect your employer to evaluate tools like this, given the sales cycles described. Learn to review AI output well. Checking a model's assumptions, spotting a broken formula link and judging whether a deck's story holds up become more valuable than raw build speed. Ask what the agent can see, what it changed, and whether you can trace every number. **If you manage a team that does diligence or reporting.** List your repetitive tasks (question lists, comparable-company tables, recurring updates) and note how many hours each takes. That inventory helps you judge any vendor pitch and decide what to pilot first. **If you are early in your career.** The learning path through manual model-building is changing. Understand how a model works well enough to audit it. The CEO said new hires learn quickly because they get "freedom to learn by failing," so look for roles where you get real responsibility. **If you are weighing a startup.** The CEO's advice was blunt: once the thought of starting a company is in your head, it is a matter of time. Take that as one founder's opinion, not a rule. ## Can you build your own diligence or deal tracker? Most teams do not need an enterprise agent platform for every problem. Plenty of finance work is smaller: a diligence question tracker that assigns each question to an owner and shows what is still open, a watchlist of companies with notes, or a request form that routes ad hoc analysis asks. [Remy](https://goremy.ai) builds this kind of internal tool from a plain-English description. It produces a full-stack app with a database, sign-in and a live URL, and the code stays in a git repository you own. You could describe your question-tracking process, with roles and statuses, and have a working app instead of a shared spreadsheet that everyone edits at once. ## Frequently Asked Questions ### What does Model ML do? Model ML builds AI agents for financial services firms. They automate work such as due diligence, M&A and investment banking tasks, and are described as monitoring companies, building LBO models and writing pitch decks. ### How much funding has Model ML raised? The CEO said the company has raised "there or thereabouts" over $150 million. That figure comes from the interview and was given approximately, not as an audited number. ### Who founded Model ML? Two brothers, the CEO and Arie, who is the technical co-founder. Model ML is their third YC company, after Fat Llama (2017) and Fancy (2020), both of which they sold. ### Why did Excel matter so much for AI in finance? Finance models live in spreadsheets. According to the CEO, AI agent tools could not interact with Excel until February 2026, which limited what they could do for analysts. Once agents could read and edit workbooks, they could take on real modeling work. ### Will AI agents replace junior bankers? The interview does not say so. It describes automating specific repetitive tasks. The likelier near-term change is that junior staff spend more time reviewing and checking AI-built work than producing first drafts by hand.