An AI agent for junior bankers' grunt work just raised $250m at a $1.5bn valuation
General Atlantic and QED Investors co-led the $250 million Series B, taking the London fintech's total funding to $350 million.
By The Graze · November 18, 2026 · 5 min read

Fintech
On 17 November, Model ML said it had raised $250 million in Series B funding at a $1.5 billion valuation, 18 months after launching from stealth. The round was co-led by General Atlantic and QED Investors and takes the London company's total funding to $350 million. The product is a personal AI agent for bankers, consultants and investors, built to finish the analysis rather than send another alert.
The raise
The Series B participants, in the company's account, were Evolution Equity Partners, Airtree, Forestay Capital, Y Combinator, FT Partners, Latitude, Inception Fund and 13Books Capital, alongside the two co-leads. Model ML said it will use the money to expand research and engineering teams in London and New York, and to support wider customer deployments. Offices opened this year in Paris, Madrid, Tokyo, Singapore and Dubai, adding to London and New York.
The company was founded by brothers Chaz and Arnie Englander, described in the announcement as repeat entrepreneurs. Chaz Englander is co-founder and chief executive. The advisory board, the company says, includes Sir Noel Quinn, former group chief executive of HSBC, and Axel Weber, former chairman of UBS.
Paul Stamas, managing director and global head of financial services at General Atlantic, called the product "purpose-built for financial and professional services at a time when these institutions are moving quickly from AI experimentation to everyday use." Nigel Morris, co-founder and managing partner of QED Investors, said: "In more than 40 years of building and backing financial services companies, I have not seen a team build faster." Eighteen months ago, he said, Model ML was emerging from stealth. "Today, its product is deeply embedded inside some of the world's largest institutions."
Those institutions are not named.
What the agent is supposed to do
In the company's account, each banker, consultant or investor gets an agent that "understands their clients, their live deals and how they like things done." It runs on the firm's own systems and data, "often" inside the customer's infrastructure, so that data need not leave the building. Customer data, Model ML said, is never used to train AI models. The company says it is ISO/IEC 27001:2022 certified and SOC 2 Type II audited.
The agent is not, in this telling, a chatbot. "AI should do more than give you another answer or notification. It should understand your work and get things done," Englander said. When news breaks on a company a banker covers, the announcement says, the agent is meant to judge whether it matters, work through the implications and prepare the analysis in the firm's own formats, "in minutes rather than days."
TechCrunch described the same product in more familiar banking language: work traditionally done by junior bankers and consultants, including building financial models, preparing pitch decks and writing due diligence reports. The startup, it said, builds agents that integrate with clients' systems and tools, including Excel and PowerPoint, and can update models and presentations when new developments change the analysis. Those task examples are TechCrunch's, not a customer case study.
Model ML said it is not tied to a single AI provider. It benchmarks models on "real financial work" and routes each task for quality, speed and cost. A growing share of that work, it said, runs on its own fine-tuned open-source models, which on the workloads they handle "run at 2% of the cost of frontier models and are, in most cases, five times faster, with accuracy that matches or beats the frontier on those tasks." That comparison is the company's own. It published no independent test and no named workload.
The figures that are missing
The growth rates are relative, not absolute. A 12-fold increase in annual recurring revenue is a strong claim if the starting point was already material, and a weaker one if it was small. Model ML did not publish the dollar figures behind either multiple. Daily active users, up 26 times year on year in the same quarter, are given without a starting or ending number. No bank, consultancy or asset manager is identified as a customer in the announcement, in General Atlantic's posting of the same text, or in TechCrunch's report. Morris's phrase about "some of the world's largest institutions" cannot be checked from the public materials.
The earlier Series A, $75 million led by FT Partners in November 2025, is confirmed in the new announcement. The company gives a total of $350 million raised. It does not, in this release, break out the earlier funding that sits between that Series A and the new total.
| Item | Figure | What is, and is not, disclosed |
|---|---|---|
| Series B, 17 November 2026 | $250 million at a $1.5 billion valuation | Official, co-led by General Atlantic and QED Investors |
| Series A, November 2025 | $75 million | Official, led by FT Partners; no valuation given in this announcement |
| Total funding | $350 million | Official; the remainder after the Series A and Series B is not itemised here |
| Q3 2026 annual recurring revenue | 12 times the year-earlier figure, the company said | No dollar figure |
| Q3 2026 daily active users | 26 times the year-earlier figure, the company said | No starting or ending count |
| Customers | Not named | QED said the product is "deeply embedded" in some of the world's largest institutions |
Compiled by The Graze from Model ML's 17 November 2026 announcement, the same text as posted by General Atlantic, and TechCrunch's report the same day. Growth rates and the $1.5 billion valuation are the company's figures. No customer names or dollar revenue were published.
Our take
The interesting fact in the announcement is not that another AI company has raised a large round. It is what Model ML says it is selling: a personal agent that sits inside a bank's systems and does the grunt work that used to train junior bankers — the model, the deck, the first pass at due diligence — before the senior person asks. If that is true at any scale, it is a change to the apprenticeship model of the industry, not just a faster research assistant. The company is careful not to put it quite that bluntly. TechCrunch is willing to. The distance between those two descriptions is worth keeping in view.
So is the evidence. A $1.5 billion valuation, a 12-fold jump in recurring revenue and a 26-fold jump in daily users are presented without a named client or a dollar of revenue. "Deeply embedded inside some of the world's largest institutions" is a memorable line from a lead investor. It is not a customer list. Banks will care, rightly, about the parts of the announcement that are easier to verify: data that stays in the client's environment, a claim that client data is not used for training, and the ISO and SOC 2 audits. Those are table stakes for this market, not proof that the agent already does the work.
The 2 per cent cost claim should be read the same way. Running fine-tuned open-source models more cheaply than frontier systems is plausible, and it is still a company benchmark on unnamed workloads. Until Model ML, or a bank that uses it, publishes a before-and-after on a real deal team, the raise tells you more about investor appetite than about how much junior work has actually moved. The appetite is clearly there. The work, for now, is described rather than shown.
Sources
- Model ML, Model ML raises $250 million at a $1.5 billion valuation, 18 months after launching from stealth, 17 November 2026
- General Atlantic, Model ML raises $250 million at a $1.5 billion valuation, 18 months after launching from stealth, 17 November 2026
- TechCrunch, Model ML raises $250M to automate investment banking grunt work, 17 November 2026







