Team members supported
The Problem
A mid-market non-profit accelerator operating across Latin America runs programmes that connect experienced mentors with early-stage companies. The quality of these matches is critical: the right mentor can transform a company's trajectory, while a poor match wastes everyone's time. Matching needs to account for company stage, geography, industry vertical, specific challenges the company is facing, and the mentor's areas of expertise and availability.
The accelerator had massive amounts of unstructured data about both mentors and companies. Mentor profiles, past engagement notes, company applications, progress reports, and programme evaluations were spread across documents, spreadsheets, and databases. The Entrepreneur Experience team of 200 people was responsible for making these matches, but they were essentially working from memory and manual searches through documents.
The result was inconsistency. Some team members knew certain mentors well and would recommend them frequently, while equally qualified mentors with different advocates got overlooked. The matching process was slow, subjective, and did not scale. As the accelerator grew, the gap between the quality of matches they wanted to make and the quality they could actually deliver kept widening.
The Solution
BetterBrain built a RAG-powered matching system that surfaces the best mentors for each company from the accelerator's unstructured data. The system ingests mentor profiles, past engagement records, company applications, and programme data, building a rich understanding of what each mentor brings and what each company needs.
When a team member needs to find mentors for a company, they describe what the company is looking for in natural language. The system uses BM25 and vector embeddings to retrieve relevant mentor profiles, then OpenAI's gpt-4o re-ranks the results based on multi-dimensional relevance: industry fit, stage expertise, geographic proximity, language capabilities, and past engagement success.
The system does not replace human judgment. It gives the 200-person Entrepreneur Experience team a powerful starting point. Instead of manually searching through documents and relying on personal memory, team members get a ranked list of the most relevant mentors with clear explanations of why each one is a good fit. The team still makes the final call, but they make it faster, more consistently, and with better information.
The Number
- 200-person Entrepreneur Experience team empowered with data-driven recommendations
- Mentor matching transformed from manual guesswork to systematic process
- Previously overlooked mentors surfaced based on actual fit, not familiarity
- Faster matching process freed team time for relationship building
- Consistent match quality across the entire team