IARA – AI Copilot for Personalized Learning

Transform the higher-education experience through a scalable generative AI learning copilot that combines personalization, academic rigor and operational efficiency. IARA was created to address critical challenges in student engagement, content overload and personalized learning, while reducing repetitive faculty work. For students, the goal was to increase autonomy through multimodal learning support, competency-based assessment, personalized leveling recommendations and academic wellbeing monitoring. For faculty, it was to automate assessment support while preserving human academic validation. At institutional level, IARA aimed to establish a secure, scalable AI capability that reduces technology costs, strengthens educational innovation and creates reusable GenAI expertise across the Ânima ecosystem.

IARA delivered measurable financial, experience and quality outcomes while establishing generative AI as a scalable institutional capability. The project eliminated BRL 1.5 million in SaaS contracts related to assessment support and avoided BRL 1 million in implementation costs for assessment solutions supporting the Competency Test and Academic Happiness Index. Veteran students participating in the 2025/1 pilot recorded NPS 6 points higher than non-participants. IARA also eliminated spelling errors and missing justifications in AI-generated exam questions. The project reached breakeven in six months, with estimated 2025 ROI above 100%. Beyond these measurable results, IARA is now used at scale across the Ânima ecosystem and has generated invitations to showcase the case at partner events, particularly national AWS events.

IARA is powered by AWS Bedrock, connecting generative AI to validated academic knowledge bases, while AWS Polly enables multilingual audio experiences. The architecture combines Generative AI, AI/ML and Big Data Analytics to deliver personalized learning, competency-based assessment, academic wellbeing monitoring and AI-assisted faculty workflows. Rather than deploying GenAI as a standalone tool, IARA integrates it with Ânima’s academic competency framework and pedagogical validation model. The technology stack was designed to balance scalability, security, cost efficiency and academic reliability for large-scale educational use. AWS, CI&T and Compass supported the initiative as technology partners.

Implementing IARA required more than deploying generative AI: it demanded the integration of technology, pedagogy, product design and data into a single scalable learning experience. Technology teams addressed AWS architecture, GenAI and integrations, while Academic teams ensured pedagogical curation, competency-framework alignment and faculty validation. Product/Innovation shaped the experience and roadmap, and Data supported NPS and usage analysis. Key challenges included rapid technology evolution, resource constraints, user adoption, high-volume infrastructure at sustainable cost, response quality and faculty enablement. These risks were addressed through wave-based rollout, continuous user feedback, market research, strategic technology partnerships, and ongoing technical and pedagogical validation, enabling IARA to progress from experimentation to ecosystem-wide scale.

Higher education faces a complex challenge: students need more personalized and engaging learning experiences while navigating an increasing volume of content and formats, and faculty spend significant time on repetitive activities such as creating and correcting assessments. Ânima saw an opportunity to use generative AI not simply as a productivity tool, but to redesign how technology supports learning.

IARA — Inteligência Artificial e Ressignificação da Aprendizagem — was created as an AI learning copilot grounded in validated academic knowledge and Ânima’s competency framework. Powered by AWS Bedrock, IARA provides students with personalized, multimodal support across text, video and audio in three languages. Capabilities include video summarization, transcription and translation, competency-based assessments with personalized leveling recommendations, and academic wellbeing monitoring through the Academic Happiness Index. For faculty, IARA generates competency-aligned questions and supports open-answer correction with personalized feedback and grade suggestions, always preserving faculty validation.

Implementation required close collaboration across Technology, Academic/Education, Product/Innovation and Data teams. The project combined AWS architecture and GenAI expertise with pedagogical curation, experience design, competency frameworks and continuous analysis of NPS and usage. A wave-based rollout and continuous user feedback allowed the solution to evolve according to real student and faculty needs. Scaling required addressing high-volume infrastructure at sustainable cost, rigorous pedagogical curation to ensure response quality, and faculty enablement. AWS, CI&T and Compass supported the initiative as technology partners.

IARA has moved into scale across the Ânima ecosystem, turning GenAI into an institutional capability rather than an isolated experiment. Beyond expanding student autonomy and reducing repetitive faculty work, the project eliminated BRL 1.5 million in SaaS contracts and avoided BRL 1 million in assessment-tool implementation costs. Veteran students participating in the 2025/1 pilot recorded NPS 6 points higher than non-participants. The project reached breakeven in six months, with estimated 2025 ROI above 100%, while building production LLM expertise that now supports future AI initiatives across the organization.

IARA is deployed in Brazil at scale across the Ânima ecosystem. Ânima has nationwide operations spanning all five Brazilian regions — North, Northeast, Central-West, Southeast and South — providing the organizational reach for scalable adoption of the solution. IARA was designed as a replicable AI model that can be expanded across educational institutions and adapted to new learning contexts, combining a shared technological foundation with pedagogical scalability.