Fyxer, an AI executive assistant, has reported significant growth, increasing its annual recurring revenue (ARR) from $1 million to $32 million in 2025. The service leverages OpenAI models to manage professional communications, such as emails, by integrating them with over 500,000 hours of real executive assistant (EA) workflows.

To handle complex conversational contexts, Fyxer employs a system of 30 to 50 specialized models rather than a single large model. These models perform specific tasks, such as classifying whether a message requires a response, analyzing intent, or retrieving relevant historical context. The system uses Direct Preference Optimization (DPO) to improve through user feedback, learning from the differences between original AI drafts and versions edited by users.

The company utilizes OpenAI's fine-tuning capabilities and Low-Rank Adaptation (LoRA)—a technique to efficiently adapt large models—to create these task-specific variants. Fyxer reported that 53% of its AI-generated drafts are currently accepted by users without modification. Fyxer's co-founder, Archie Hollingsworth, stated that the company aims to expand the assistant's capabilities beyond drafting toward managing broader communication and coordination workloads.

Sources: