evidoria

← Back to browse

Good practice Imported

Pref Rio — Rio de Janeiro's AI WhatsApp Agent Proactively Connects Vulnerable Families to Public Services

Brazil · Rio de Janeiro · See the Brazil profile · See the Rio de Janeiro profile

Evidence: Descriptive / self-reported Top 47% 60/100 · Ask Evidence Copilot about this practice

IplanRio's 'Pref Rio' AI agent proactively messages vulnerable Rio de Janeiro residents on WhatsApp to connect them to health, education and welfare services; the city reports higher school attendance and benefit-renewal rates among the 135,000 residents initially targeted.

135,000
Initial target population reached
10 %
Increase in school attendance in lower-income neighbourhoods
25 %
Previously dropped-out students returning to school
7 percentage points
Increase in Bolsa Família benefit renewals
1 US$ million
Bloomberg Philanthropies Mayors Challenge grant

Details

Maturity
Pilot
Promoter
IplanRio (Empresa Municipal de Informática, Rio de Janeiro City Hall)
Period
May 2026 – present
Keywords
social services, education, health, citizen engagement, WhatsApp

Context

Rio de Janeiro's municipal technology company IplanRio, linked to the mayor's Civil House, launched 'Pref Rio', an AI agent on WhatsApp, on 8 May 2026. The tool cross-references municipal databases with individual WhatsApp conversations to identify residents' needs and connect them proactively to public services such as healthcare, vaccination campaigns, school enrolment and attendance support, and social-benefit programmes.

Activities

The initial phase targets around 135,000 people in situations of greater vulnerability, focused on pregnant women and children up to six years old, with planned later expansion to elderly residents, people with disabilities, women, and eventually the wider population. The project was one of 24 winners worldwide of the Bloomberg Philanthropies 2025–2026 Mayors Challenge, each receiving US$1 million plus operational support and funding for dedicated staff.

Results

City officials report early-testing results: roughly a 10% increase in school attendance in lower-income neighbourhoods reached by personalised WhatsApp messages, about 25% of previously dropped-out students returning to school after digital contact combined with Department of Education follow-up, and a 7-percentage-point increase in Bolsa Família benefit renewals among supported families.

Conclusions

These figures are city-reported early-testing results rather than an independently audited evaluation, and the eligibility and matching logic behind the AI agent has not been published.

Implementation

Indicative cost
Low (< €50k) — Funded via a US$1 million Bloomberg Philanthropies Mayors Challenge grant plus operational support and dedicated staff funding.
Time to results
Short (< 1 year) — Launched 8 May 2026 targeting around 135,000 residents, with a planned phased expansion to additional vulnerable groups and eventually the wider population.
Staffing & skills
IplanRio (Empresa Municipal de Informática) technical team, Department of Education follow-up staff, Bloomberg Philanthropies-funded dedicated scaling staff

Conditions for success

  • Integration of municipal databases with WhatsApp messaging
  • Bloomberg Philanthropies Mayors Challenge funding and operational support
  • Phased rollout starting with the highest-vulnerability groups

Common failure modes

  • Eligibility and matching algorithm not published
  • Results are early-testing, city-reported figures rather than independently audited

Where it fits

Governance type
municipal government
Scale
city
Income level
upper-middle-income

Commonly funded by

Philanthropic / foundation funding Own resources / municipal budget

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Similar practices you may find useful