According to OpenAI, Fyxer has developed an AI executive assistant designed to organize inboxes and draft emails by following conversations across multiple tools. Rather than relying on a single text-generation model, Fyxer divides email management among 30 to 50 specialized models handling tasks like classification, intent analysis, retrieval, and email generation. The company utilizes frontier OpenAI models and fine-tuning capabilities, drawing upon a dataset built from more than 500,000 hours of historical human-powered executive assistant workflows. Fyxer applies supervised fine-tuning and Low-Rank Adaptation for task-specific variants, and uses OpenAI's fine-tuning platform and managed fine-tuning team. The system continuously improves through real user feedback by converting edits made to drafts into training data using Direct Preference Optimization. Every drafting change undergoes A/B testing before deployment. Fyxer reported that 53 percent of its AI-generated drafts are accepted as written. Additionally, the company grew from $1 million to $32 million in annual recurring revenue in 2025. Moving forward, Fyxer aims to expand its assistant's capabilities to manage broader communication and coordination workloads for users.
How Fyxer built an AI executive assistant people trust
Fyxer has built an AI executive assistant that uses OpenAI models and fine-tuning to manage inboxes and draft emails. The system relies on dozens of specialized models, user feedback, and historical executive workflow data. Fyxer reported growth in annual recurring revenue from $1 million to $32 million in 2025.
Why it matters
Fyxer's approach demonstrates how combining specialized model architectures, domain-specific training data, and automated preference optimization can scale a contextual AI product to millions in revenue.
Topic dossiers OpenAI