Connect Lead / Senior Delivery Consultant

Comsysto Reply has partnered with Audi to develop an AI-based multi-agent system for the carmaker's Used Car Platform, a digital marketplace for B2B used car trading. Since 2025, the platform has accounted for nearly 100% of Audi's B2B business in this sector. The solution uses three specialised AI agents built on GitHub Copilot to handle requirements analysis, code generation, and quality assurance. The agents work with a dedicated knowledge base containing the platform's architectural decisions and specifications, enabling them to produce outputs consistent with existing systems. The system follows a human-in-the-loop model, with AI handling repetitive tasks whilst developers focus on review and validation. Comsysto Reply has supported the platform since 2021, beginning with migration from legacy systems to AWS-based architecture.

Reply has released a new set of Prebuilt AI Apps, production-ready agentic applications designed to accelerate enterprise AI adoption. The ready-to-use solutions combine process knowledge, curated datasets, domain ontologies and agentic workflows to deliver immediate operational value. The applications address key enterprise areas including credit evaluation, compliance assessment, manufacturing intelligence, HR processes, procurement and content production. They can be customised and integrated with existing enterprise systems whilst maintaining governance controls. Reply's Prebuilt AI Apps aim to help organisations transition from AI experimentation to scalable deployment by providing structured starting points that reduce initial complexity. The solutions automate recurring activities, improve decision-making through structured context and enhance user experience via conversational interfaces. The applications support industries including banking, critical infrastructure and manufacturing.

Reply has launched Model Factory, an industrial production line for building generative AI models grounded in corporate knowledge. The platform enables organisations to train AI models using their internal documentation, software repositories, business data and domain expertise within secure vaults. The factory industrialises data preparation, training, evaluation and deployment through a controlled lifecycle. Its training layer includes pre-training for domain awareness, supervised fine-tuning for specific tasks, reinforcement learning for policy alignment, and distillation for efficient deployment. Each model is released as a versioned enterprise asset with quality controls and EU AI Act compliance. Reply CEO Tatiana Rizzante said proprietary models will become key strategic differentiators, allowing companies to transform distinctive assets into governed AI models that remain under their control whilst enabling continuous improvement and specialisation.