Bridgetown Research raises $19M to speed up due diligence with AI | TechCrunch

Due diligence is a costly business, and not just in the realm of investing. Even for a company trying to launch a new product or explore a partnership, finding the right data and doing the research can take weeks and get very costly if they are to make an educated decision — especially when third party agencies and consultants get involved.

A new AI startup called Bridgetown Research says it can make a dent in that cost base and speed up the process by using AI agents that can do most of the data collection and research work that goes into due diligence. And as part of this effort, the startup recently raised $19 million in a Series A round co-led by Accel and Lightspeed.

Co-founded in December 2023 by its CEO Harsh Sahai, a former McKinsey employee and research scientist at Amazon, Bridgetown Research has built three types of AI agents that it claims can gather information, collate and condense that data, and finally present it in an easy-to-read format.

Bridgetown is exploiting the very networks that consultants and researchers often use to gather insights: networks of industry experts who can provide insights on a particular company or sector. The startup essentially partners with these expert networks, and then uses its AI voice agent to interview experts for the information the company needs to find for its clients.

“Because insiders don’t have to schedule a call with a human being, they can log on whenever and have a conversation,” Sahai said in an interview. “Instead of talking to one senior executive, you can talk to mid-tenure people, but a lot more of them… at a much higher scale.”

Bridgetown’s second set of agents then use large-language models (LLMs) alongside tools for clustering and regression to interpret the data collected by the voice agents, and pass this information back to the LLMs to summarize the answers. Finally, the third set of agents uses small-language models to reproduce the interpretation in a digestible form, like a presentation.

Using these agents, the startup says it can produce an initial due diligence analysis in 24 hours with inputs from hundreds of respondents.

Sahai said clients can either use Bridgetown’s agents to gather data and insights on their own, or they can hire an independent consultant or a small consulting firm to work with the agents to get the same quality of analysis as they would from firms like McKinsey or Bain.

That sounds appealing, but large language models and the AI agents built on top of them still tend to hallucinate — they tend to just make up information. So how is an investor to trust research reproduced by an AI agent? Sahai says the startup addresses this with its “steerability and audibility” approach.

This means, he explained, clients can review the data and trace every step the agent took to arrive at its conclusions, similar to the “reasoning” AI models out there. Additionally, the voice agents record their conversations with the experts they interview so that the information can be manually verified.

He added that the AI agents do not rely on a single data source. Instead, they gather information from multiple sources, interpret it using large language models, and then employ fine-tuned models to process the data.

“We haven’t seen our approach before,” Sahai said. “Most platforms leave it to you to collect the information you need, and then they will process it on your behalf.”

Bridgetown isn’t the first to tackle this opportunity to make due diligence easier — we already have startups like Mako AI and DiligentIQ in the space. However, Sahai thinks other platforms do not provide a complete enough solution.

Bridgetown Research has two customers in the U.K. and a dozen in the U.S. These include top-tier private equity and venture capital funds, consulting firms, and big corporations that address the M&A pipeline, Sahai said.