Fyxer Transforms Communication Management with AI-Powered Assistant
Fyxer has developed an AI-based executive assistant that manages professional correspondence using OpenAI models and has earned the trust of users.
Fyxer has launched an AI assistant trained on OpenAI models and thousands of hours of executive workflows to track conversations and commitments in business processes.
Communication Challenges in Business Processes
For professionals, working life requires tracking conversations and commitments across invoices, meetings, messages, and apps. Without solid context, these commitments can be missed, harming projects and relationships.
Fyxer designed an AI-powered executive assistant that tracks this process as business flows across different tools. This system combines the latest OpenAI models with a vast number of executive workflows.
The Role of AI in Email Management
Email is one of the clearest areas where this technology is seen. Two people can receive the same email, but depending on the relationship, past history, and desired goal, they may need to respond completely differently.
This makes a seemingly simple task quite difficult for AI, and Fyxer solves this complexity by learning each user's working style.
Moravec's Paradox and the Assistant's Working Principle
Archie Hollingsworth, Co-Founder of Fyxer, highlights Moravec's paradox, which states that things that are easy for humans are hard for computers, and things that are easy for computers are hard for humans.
The system leverages frontier OpenAI models to support the core part of the experience, from understanding the email to finding the right context and drafting.
Reasons for Choosing OpenAI Models
Fyxer chose this technology because OpenAI models perform best in internal benchmarks and offer strong fine-tuning capabilities for subjective tasks like tone and intent.
Additionally, hands-on engineering support was provided throughout the process via whiteboard sessions and technical collaboration.
Key Lessons for Developing AI Products
Fyxer's approach offers strategic lessons for founders building high-context AI products, such as breaking email down into smaller tasks.
The system is built around thirty to fifty customized models responsible for a narrow slice of the email workflow.
Training with Real Data and the Feedback Loop
The company spent years running a human-backed executive assistant service, accumulating a dataset of over two hundred thousand hours of annotated executive workflows.
Fyxer's system continues to improve through real user feedback using Direct Preference Optimization.
Growth and Success Metrics
Today, fifty-three percent of AI-generated drafts by Fyxer are accepted as written.
In 2025 alone, Fyxer grew its annual recurring revenue from one million dollars to thirty-two million dollars, while achieving a user retention rate of over ninety percent in a ninety-day period.