V7 Enhances AI Agents with Institutional Memory Using OpenAI Models

V7 has achieved 89% accuracy on the hardest graph-query tests with its integration of OpenAI's GPT-6 Astra. Co-founders Alberto Rizzoli and Simon Edwardsson emphasize that AI must learn business operations as effectively as it learns from the Internet to tackle complex enterprise use cases in finance and insurance. V7's agents can now complete workflows with 99.9% accuracy and reduce deal screening time from 15 minutes to under a day, resulting in a 21x speed increase for asset managers. The implementation of GPT-5.6 Luna has led to an 11.6-point increase in accuracy over previous models, while GPT-5.6 Sol has cut tool-call error rates from 2.7% to 0.2%. V7's use of OpenAI's models has also resulted in a 78% reduction in cost per document compared to GPT-5.4 mini, and a 5% decrease in token usage for PDF-heavy workflows. This follows V7's founding in 2018 and reflects a strategic shift towards enhancing AI capabilities for mission-critical tasks.