Danny Goldman was up particularly late one night in 2021 as he worked on a deal regarding a company that sold technology to parking garages.
As a former Bain and Co. consultant working with private-equity firms, he was responsible for compiling data from 25 customer calls he took as part of his due diligence process. He spent five hours packaging the customer data and still had to compile a preliminary market analysis on the total addressable market. At 2 a.m., he found himself trying to tap Bain’s institutional data about similar companies, but there was no way he could conduct the search.
Though Goldman “totally struck out” that night, hitting the wall would give him the idea for a startup idea that is coming to fruition.
On Tuesday, Goldman and his cofounder, Shivaal Roy, launched Mako AI after building the startup for about a year. Alongside the launch, Mako also announced its $1.55 million seed round led by Khosla Ventures.
The startup offers a Wall Street-focused generative AI agent that pulls on a firm’s enterprise data to answer questions and conduct complex analyses. It aims to solve the woes of many early-career private equity associates, like collecting data, writing reports, and analyzing companies.
“I was spending three to five hours a day looking for information, doing rote synthesis, building formulaic output, and it was very clear that something was pretty broken here,” Goldman, CEO of Mako, said. “I started to talk to more people that had similar jobs and it was clear that this was kind of a universal truth,” he added.
Letting employees search institutional data has traditionally been tricky because of complexities with user permissions and data quality, Roy, Mako’s CTO, said. But that hasn’t stopped some of the biggest PE firms from trying to improve enterprise search. Blackstone, for example, built its DocAI platform for AI-powered search. At the same time, KKR developed RealHouse for its real-estate teams to centralize portfolio and deal data in one place to find data almost instantly.
How Mako works under the hood
Mako uses large-language models, including OpenAI’s ChatGPT, to string together a network of multiple AI agents that can tackle different parts of a given task. Also important is the knowledge graph that tells the models which documents are best suited to answer specific questions. Mako conducts this ranking of different documents during the 30-minute onboarding process, wherein other AI models are used to read documents and have context on which ones are good for different types of information, Roy said.
If a user wanted to know what customers think of a given company, the knowledge graph would probably point the models to customer-call data, but if someone wanted to know more about revenue or customer count, that would likely be in the pitch deck.
Before cofounding Mako with Goldman, Roy was an early engineering hire at Glean, an AI-enabled enterprise-search startup valued at $2.2 billion. He said that some components of Mako’s tech are built similarly to Glean.
Mako has also been trained on specific workflows common in private equity, like benchmarking companies against each other, drafting sourcing emails, or writing investment committee memos, Goldman said. Each output or sentence, whether it’s a simple answer to a question or a complete analysis, is cited back to the specific document, which helps mitigate hallucinations, Roy added.
So far, Mako has a handful of clients, including mid-market PE firms and growth-equity firms, including L.A.-based Shamrock Capital and San Francisco firm GroundForce Capital. Mako, which has one other full-time engineer, plans to build its engineering organization with the fundraise.
“The most important thing right now is getting this product from a first-year associate to a second-year associate,” Goldman said.
Here’s the pitch deck Mako used to raise its $1.55 million seed round led by Khosla Ventures.
Share
Print