Initiative · Multi-country · 2017 — ongoing

ConsiliumBots

Using data and technology to help people make important decisions

A non-profit using data and technology to help people make important decisions — chatbots, smart matching platforms, and personalized guidance across education, housing, and benefits. 36 tools, 6M+ people, 7 countries.

New York · Santiago

Digital guidance, built for real policy environments

I co-founded ConsiliumBots with Felipe Saint-Jean. ConsiliumBots is a non-profit that partners with governments and researchers to design, build, and evaluate digital guidance tools — chatbots, smart matching platforms, search interfaces, and personalized communication — that help people navigate high-stakes decisions in education, housing, and access to public benefits.

To date the team has shipped 36 tools reaching 6M+ people across 7 countries (Chile, Colombia, Ecuador, Peru, Brazil, the Dominican Republic, and the United States), working directly with national ministries of education, central banks, the Inter-American Development Bank, and U.S. school districts.

What we build

  • AI-powered chatbots that reduce information frictions in admissions and benefit-take-up
  • Smart matching platforms for school choice and teacher assignment (CCAS)
  • Personalized feedback for students and teachers based on assessment results
  • Coordinated choice and assignment systems used by ministries at national scale
  • Field-experiment infrastructure so every product can be evaluated rigorously

Selected programs

  • Decidiendo Para un Futuro Mejor (DFM) — information interventions for college and major choice
  • Smart Matching Platforms — teacher assignment optimization in Ecuador and Peru
  • Subsidy Project — increasing take-up of education and family benefits
  • Feedback for Teachers / Students — closing the loop between assessment and classroom practice

A flagship application is the Colombia higher-education guidance work: nationwide, at-scale experiments embedded in the ICFES results portal tested which information interventions actually change information-seeking behavior, and what does (and does not) transport when you scale.

The broader goal is simple: combine user-centered product design with rigorous measurement so that public information systems can both deliver better guidance and learn quickly which designs are worth scaling.

Visit ConsiliumBots ↗ Related work: adaptive experimentation at scale ↗