The initiative had reached a critical scale inflection point, requiring stronger operational governance, enterprise-grade architecture, and production readiness to support 46 million citizens with reliability and trust.

PS LATAM and the client co-led a production hardening program: architecture redesign, MuleSoft V3 integration modernization, controlled rollout governance, and performance optimization that reduced response time from 36 seconds to 5 seconds. Poupatempo scaled from 2,500 conversations/day at launch (Feb 23) to a steady 20,000 conversations/day (Apr 2026), achieving 15.1% scheduling conversion vs. 7.9% on legacy channels.

On May 8, 2026, Buscador (Motor IA) went live on poupatempo.sp.gov.br, the first conversational AI search agent deployed in a Brazilian state government portal, serving 220K+ daily requests across a 40M+ citizen base. Two additional use cases (SEI and Zeladoria) have POCs finalized and are pending stakeholder presentation.

The platform serves 46M citizens with enterprise-grade reliability, doubling conversion rates and maintaining sub-1% error rates. Competitively, PS LATAM is the preferred partner for the SEI use case where Microsoft has not succeeded. The program influenced $25M+ in pipeline (SECOP 2025), and a joint Prodesp+Detran workshop (Apr 28) signals further public-sector expansion across São Paulo State.

Claro needed to elevate service consistency and speed across 20,000 agents while reducing operational dependency on retraining cycles and fragmented knowledge management.

Salesforce and Claro implemented an AI-enabled knowledge operating model through VIKI, centralizing service intelligence and deploying real-time guidance to all agents. The model is now extending to AIDA for customer-facing orchestration.

Operationally, service quality became more standardized and knowledge deployment accelerated across centers. Strategically, Claro now has a scalable AI service operating model with regional replication potential across LATAM operations.

Dataprev and Salesforce jointly designed an AI operating model to support citizen services at federal scale - balancing speed, reliability, governance, and trust across multiple ministries and high-volume citizen journeys.

A governance-first multi-agent architecture was co-created: ministry-specialized agents, a router for orchestration, RAG constrained to official datasets, and observability scoring for continuous control. This established a trusted and reusable AI framework for federal operations.

Operationally, Dataprev enabled federal-scale AI interactions with controlled quality and reduced response times for citizens. Strategically, it created a reusable governance model for expansion across ministries. Commercially, it strengthened Salesforce positioning for public-sector AI, Data Cloud, and integration growth.

iFood needed to transform partner acquisition from a labor-intensive sales motion into a scalable digital growth engine, serving 55M app users and processing 120M orders/month across Brazil.

With Diana Inbound and Diana Outbound on WhatsApp, iFood and Salesforce implemented an AI-driven acquisition operating model, from lead engagement to digital contracting, fully replacing Blip, the previous competing platform, and achieving 100% migration. Diana 3.0 introduced a redesigned architecture with multi-agent integration, image recognition pilot, and voice pilot in progress.

Diana consistently exceeded signed contract targets, with Diana Outbound delivering 2× the monthly target in contracts. 3-week average consumption reached 1.13M actions (+53% WoW delta). Performance gave iFood the confidence to approve an additional $2M in Flex Credits and graduate the FDE Momentum Program in April 2026.

Agibank needed to reduce high-volume service friction while converting customer service moments into proactive revenue opportunities through a unified customer intelligence model.

Agibank and Salesforce deployed coordinated service and upsell agents powered by Data Cloud 360, enabling autonomous resolution for routine intents and context-aware cross-sell recommendations during the same customer journey.

Operationally, autonomous resolution reduced service load and improved efficiency. Financially, cross-sell opportunities increased through AI-enabled service channels. Strategically, the bank established a scalable AI banking operating model where Data Cloud 360 acts as a revenue and experience multiplier.