Nubank Develops Framework for Customer Support AI Agents Serving 100M+ Users
Researchers from Nubank published a framework for building and deploying customer support AI agents at scale, reporting a 37 percentage-point improvement in AI transactional Net Promoter Score in one deployment. The system integrates structured context engineering, human-in-the-loop prompt iteration, and LLM-based evaluation to bridge offline development with live production outcomes. The work is notable for demonstrating that rigorous offline evaluation metrics reliably predict real-world performance, potentially accelerating how quickly AI agents can be safely deployed in high-stakes customer-facing roles.
A team of researchers at Nubank, a Brazilian fintech serving over 100 million users, has released a preprint on arXiv describing a unified framework for deploying large language model (LLM)-based customer support agents in production. The framework combines structured context engineering, systematic human-in-the-loop prompt refinement, and a rigorous LLM-as-judge evaluation pipeline with measured inter-rater agreement and GEPA optimization for consistency. The authors report results across five distinct deployment domains: card delivery, debt management, credit-limit support, card management, and product explanation. In the card-delivery use case, large-scale A/B testing showed a 37 percentage-point improvement in AI transactional Net Promoter Score and a 29 percentage-point gain in self-service rate compared to prior agent variants. A key finding is a strong correlation between offline simulation metrics and online outcomes, suggesting that high-quality evaluation pipelines can reliably predict production impact and accelerate iteration velocity. The paper also reports that AI customer satisfaction reached within a few percentage points of expert human agents on most tested use cases. The work was submitted to arXiv on June 7, 2026, and is linked to a related ACM publication.
What's missing
Long-term customer satisfaction trends beyond the reported A/B test windows are not discussed, nor are cost comparisons between AI and human agents.
What different sources said
- arXiv cs.CLCenter
Building Customer Support AI Agents at 100M-User Scale: An Evaluation-Driven Framework
Related
Gut Bacteria Enzyme Found to Break Down Heat-Processed Food Compounds, Producing Novel Biogenic Amines
Researchers have discovered that an enzyme in common gut bacteria can degrade N-epsilon-carboxymethyllysine (CML), a compound formed during thermal food processing, producing previously unknown biogenic amines. The enzyme, ornithine decarboxylase SpeC from enterobacteria, acts on CML and related modified lysine derivatives through a low-level 'underground' catalytic activity. This finding suggests a previously unrecognized communication axis between thermally processed dietary compounds and gut microbial physiology, with potential implications for host health.
Full-Length Gene Sequencing Reveals Two Distinct Bacterial Communities in Black-Legged Ticks Expanding Into Canada
Researchers used Oxford Nanopore full-length 16S rRNA gene sequencing to characterize the microbiome of Ixodes scapularis black-legged ticks collected in Nova Scotia, Canada, distinguishing between tick-adapted bacteria and environmentally acquired bacteria. The study comes as I. scapularis — the primary vector of Lyme disease — is rapidly expanding northward into Canada due to climate change. The findings suggest that environmentally derived bacteria in tick microbiomes are not mere contamination, which has implications for how tick microbiome data is collected and interpreted across surveillance studies.
Study Identifies Metabolic Link Between Cell Envelope Stress and Biofilm Formation in Bacteria
Researchers have discovered that the metabolite acetyl-CoA directly inhibits enzymes that degrade the bacterial signaling molecule c-di-GMP, connecting cell envelope biosynthesis stress to biofilm formation in Pseudomonas aeruginosa. The study found that sub-inhibitory concentrations of antibiotics targeting early peptidoglycan biosynthesis — but not other antibiotic classes — elevate c-di-GMP levels by reducing phosphodiesterase activity, with acetyl-CoA competing for the enzyme active site. Because the relevant enzyme domain is broadly conserved across bacterial species, this checkpoint mechanism may be widespread and could have implications for understanding antibiotic-induced biofilm responses.