How Brian Impact Foundation Uses Claude to Review 100x More Fellowship Candidates
Brian Impact Foundation, a Korean nonprofit supporting social innovators through grants and fellowships, deployed Claude to build a candidate evaluation platform called BEES (Benevolence Enhancing Expert System). With a team of just 13 people managing millions in grants, Claude enabled them to review 20,000 fellowship candidates in a single year—100 times more than the year prior—with three times as many reviews per candidate.
Impact
100x (20,000 vs. 200)
Increase in fellowship candidates reviewed
3x increase
Reviews per candidate
Challenge
Brian Impact Foundation’s 13-person team could only review a fraction of potential fellowship candidates using manual research and expert network referrals, limiting the equitability and reach of its selection process for social innovators.
Solution
Claude was integrated into BEES (Benevolence Enhancing Expert System), a custom platform that uses Claude to gather open-web data on potential candidates, extract and summarize candidate information, analyze data quality, and organize profiles by evaluation criteria—enabling reviewers to conduct thorough assessments at a scale of 20,000 candidates per cycle.
Tools & Technologies
What Leaders Say
“Partnering with Anthropic has helped our team of 13 discover a larger number of fellowship applications and review every application more thoroughly. This enables us to increase equity in our fellowship application process, strengthening our mission to support innovators using technology to make the world a better place.”
“Claude’s long context window was significant for us because we process a lot of news articles.”
Sign up to read complete case studies, access detailed metrics, and unlock all use cases.
Full Story
Brian Impact Foundation was created to find and fund the next generation of social innovators—people using technology, and increasingly AI, to solve global problems. Founded after Kakao’s Brian Beom-su Kim pledged more than half his net worth to social causes, the organization manages millions in grants with a team of just 13 people. Its fellowship program identifies and supports individual innovators, but identifying the right candidates from a global pool while operating lean was a persistent constraint.
The core problem was volume and equity. Without AI, the team relied on its expert network to surface candidates, leaving a vast field of potentially deserving innovators unreviewed. Reviewers had to conduct laborious manual searches to compile candidate profiles, and with limited staff hours, the fellowship effectively had a ceiling on how many people it could fairly consider.
The team built BEES—Benevolence Enhancing Expert System—on Claude. BEES gathers diverse data related to social problems and innovation from the open web, then sends the raw data to Claude to extract candidate information, summarize it, and analyze data quality and structure. Claude then organizes the information according to the foundation’s evaluation criteria and surfaces it in a custom web app, where reviewers can access a comprehensive view of each prospective candidate. Claude’s long context window was a decisive factor, given the volume of news articles and unstructured source material processed for each candidate.
The outcome was a step change in scale. In the most recent fellowship cycle, the foundation reviewed over 20,000 candidates—100 times more than in the prior year. Claude also tripled the number of reviews per candidate, enabling the team to apply more thorough evaluation to every applicant rather than skimming the surface of a smaller pool. The process became more equitable: the foundation could support innovators it would never have reached through its prior network-dependent approach.
For Brian Impact Foundation, AI is not just an operational lever—it’s central to the mission. By automating the data-gathering and summarization work that consumed reviewer time, Claude freed the team to focus on judgment and impact. The foundation can now pursue a larger, more diverse candidate pool while maintaining its strict commitment to evaluating every application thoroughly.