Fireflies Exposed: Billion-Dollar “AI Bot” Startup Used Human Note-Takers Behind the Scenes
Fireflies, a startup now valued at $1 billion, built its brand around an AI-powered bot that could automatically join meetings, listen to conversations, and generate structured notes. But according to co-founder Sam Udotong, the company’s earliest version relied on something entirely different — two real people quietly joining calls and taking notes by hand.
Early “AI” Was Actually Two People Writing Notes Manually
Udotong revealed the story on LinkedIn, describing how the company used human note-takers to meet early customer expectations. These workers joined conference calls, listened in real time, typed summaries, and sent them to clients within roughly ten minutes — mimicking the expected behavior of an automated transcription bot.
The approach was a stopgap solution, designed to keep the product alive while the team struggled to build a functional AI system. Udotong described it as a necessary step for survival during what he called “months of barely scraping by on pizza and determination.”
A Billion-Dollar Valuation Built on Grit and Improvised Solutions
Fireflies’ current valuation shows how far the company has come since those improvised early days. The note-taking platform has evolved into a full-fledged AI productivity tool, with automated transcription, summarization, and workflow integration. But the revelation highlights the often messy reality behind early-stage AI startups, where marketing outpaces engineering and teams improvise to meet user demand.
According to Udotong, the company endured six failed attempts to build its intended AI model. The repeated setbacks nearly killed the project — until the founders decided to rely on human operators to simulate the system’s intended functionality while they continued rebuilding the technology from scratch.
A Common Pattern in Early AI Startups
Industry analysts note that Fireflies’ strategy echoes a broader pattern in the AI sector. Many companies launch with “Wizard of Oz” systems — customer-facing automation supported on the backend by human labor. The goal is to validate demand before investing heavily in full-scale AI development.
While not inherently unethical, the practice raises questions about transparency, especially when companies market features as fully automated. Fireflies’ disclosure now provides a rare public look at how such systems often begin.
Building Real AI After Six Failed Attempts
Udotong emphasized that the team eventually built its own functioning AI engine after months of trial, error, and rebuilding. The early human-powered solution bought the founders the time and revenue they needed to continue development. Today, Fireflies promotes itself as a genuine AI productivity platform used by thousands of businesses globally.
The CTO’s story highlights the emotional toll of the process, recalling long nights, technical dead ends, and a period he described as “six failures separated by slices of cheap pizza.”
Conclusion
Fireflies’ journey from human note-takers to a billion-dollar AI company illustrates the chaotic, scrappy nature of early AI innovation. While the revelation may spark debate about transparency in the industry, it also underscores how real-world startups often survive — not through flawless technology from day one, but through persistence, improvisation, and a willingness to build the product by any means necessary until the actual AI is ready.
Editorial Team — CoinBotLab