TLDR
A leading VC firm leveraged deepset Cloud to transform their data repository into an AI-powered legal search tool, reducing research time from hours to minutes. Completed in 2.5 months, the RAG-based solution achieved 60% team adoption and now serves as their daily co-pilot.
Key Metrics:
- Research time reduced from hours to minutes
- Development speed increased by 100x compared to building from scratch
- 60% adoption rate among team members
- AI-powered tool developed and deployed in just 2.5 months
In the fast-paced world of venture capital, information is power. One leading VC firm found itself sitting on a goldmine of proprietary data, but lacked the tools to effectively capitalize on its value. This case study explores how they partnered with deepset to build an AI-powered tool for their legal department in less than three months, setting a new standard for efficiency in the industry.
The Challenge: Turning Data into Actionable Insights
Years of investment had left the firm with a vast repository of valuable information spread across departments that comes in multiple file types. This data had immense potential to provide unique insights and a significant competitive advantage. However, its confusing file structure and sheer volume made it difficult to access and use effectively.
The company's AI project lead recognized the untapped potential: "This data is one of our greatest assets. Having all this information is amazing – but only if you can use it.”
Previous attempts to organize and access the data more easily with enterprise search solutions proved unsatisfactory. "A lot of enterprise search tools promised a lot and didn't deliver much," she explained. "They expected your data to be in a certain format. Either you did a lot of heavy lifting, or it didn't work."
Plus, she knew that what the firm really needed wasn’t just a one-time implementation, but the starting point for a comprehensive AI strategy that could be scaled across the enterprise.
Choosing the Right Solution and Pilot Project
With deepset, the company saw an opportunity to move beyond traditional enterprise search tools. The deepset platform enables AI teams to build custom LLM-powered products for a variety of use cases - quickly and securely. Its guided workflows take teams from idea to launch, following best practices for AI product development.
The tech team began prioritizing projects across the company. They quickly identified the legal department as having the biggest pain point: Finding answers in a large document storage blob. "Legal is constantly searching for the right documents. They were spending at least 20 percent of their time – every single person, every single day – looking for answers to questions."
The AI solution was designed to provide the legal team with well-crafted answers to natural language questions, complete with references to the underlying data. This would dramatically reduce research time, allowing the team to spend more time on more substantive and meaningful tasks.
Partnering with deepset: From Concept to Production
The development process consisted of three key stages:
Prototyping (8 weeks)
- Starting with a small, non-sensitive subset of the data – about 100 supplier contracts
- Received weekly guidance from deepset's AI engineers
- Developed a prototype that impressed the legal team
Refinement and SME Feedback (3weeks)
- Scaled to the entire document repository
- Leveraged deepset's API for seamless integration
- Incorporated feedback from legal subject matter experts
Deployment to Production
- Launched a robust tool ready for wider use
- Completed in 2.5 months from concept to production
The project lead praised deepset's ability to accelerate the AI development process: "I can tweak data, test prompts, and spin up fast prototypes without having to be highly technical. It's 100 times faster than building from scratch.
Immediate Impact
The tool consists of a Retrieval Augmented Generation (RAG) setup that provides answers to legal questions based on the firm's proprietary data. Annotated references to the underlying documents allow the legal professionals to double-check a source for verification. The impact of this new AI-powered system is significant:
- Tasks that once consumed much of the workday are now completed instantly.
- Legal professionals can easily access the information contained in the data, regardless of where it lives.
- 60% of team members have adopted the new tool as a full-fledged co-pilot for their daily tasks.
- Other teams are requesting similar tools for their departments.
Lessons Learned and Future Plans
Encouraged by the success of the pilot project, the firm is now exploring:
- Co-pilots for other departments, leveraging the modularity of deepset – the ability to extend an existing setup and customize it to perfectly fit different use cases.
- Setups that leverage visual QA capabilities to analyze pitch decks, charts and tables for the investment team
The deepset difference
deepset enabled the VC firm's AI team to quickly and effectively build, refine, and deploy a sophisticated LLM-powered tool, thanks to the platform's capabilities for
- Rapid prototyping and iteration capabilities.
- Flexibility to customize solutions for specific use cases.
- Support for a wide range of models and system configurations.
- Built in AI development best practices.
The project lead emphasized the importance of finding a platform that bridged the gap between out-of-the-box SaaS and a fully self-hosted solution from scratch: "I needed a platform in the middle that offered fast time to value and simply allowed us to start from templates and customize the last 20 to 30 percent that we need." This approach fit perfectly with the company's vision of embedding AI throughout the organization.
By partnering with deepset, the VC firm not only solved its immediate challenges, but also gained a competitive edge in the AI-driven future of venture capital.
Get started on your ai journey today.
Table of Contents
