.svg)
Fireflies.ai is now one of the most used AI meeting assistants in the world.
Millions of users. Hundreds of thousands of business customers. A valuation that exceeds one billion dollars.
However, in its early days, Fireflies did not rely on any artificial intelligence.
No model. No automation. No advanced technology.
Just two founders... taking notes by hand.
This story, as disturbing as it is instructive, poses a central question for the current AI ecosystem:
Where does product validation end, and where does the technological lie begin?
Fireflies.ai in 2017: a promise before technology
Fireflies.ai was founded in 2017 by Sam Udotong and Krish Ramineni, in San Francisco.
The promise is simple and ambitious:
- an AI assistant joins your meetings,
- listen to the exchanges,
- take notes,
- and automatically sends a report.
Today, this type of tool is common.
But in 2017, the context is very different:
- no LLM,
- no reliable speech recognition on a large scale,
- no ready-to-use infrastructure.
The technology that is needed simply does not exist.
“Fred from Fireflies”: the AI that wasn't an AI
Behind the AI assistant called “Fred”, there was no automation.
Fred, it was:
- Sam and Krish,
- taking notes manually,
- during more than 100 meetings,
- in total silence.
Two bankrupt founders, six entrepreneurial failures behind them, nothing to lose.
The principle of “Fake It Until You Make It”, taken to its extreme.
Today, a very real unicorn
The contrast is brutal.
Today, Fireflies.ai is:
- More than 20 million users,
- More than 500,000 corporate customers,
- one valuation greater than $1 billion.
The product is real. The technology exists. The automation is effective.
But the path to get there is profound question.
Fraud or genius?
This case is divisive.
On one side:
- accusations of fraud,
- deception about technology,
- false marketing.
On the other:
- admiration for resilience,
- intelligent validation of a need,
- gradual execution.
The reality is more uncomfortable.
👉 Yes, Fireflies lied about the technology.
👉 No, Fireflies did not lie about the problem to be solved.
And that's exactly where the debate gets interesting.
Product validation vs technological lie
Validating a use before investing heavily in technology is a healthy practice.
Testing a real need avoids building a useless gas plant.
But selling a Non-existent AI, even temporarily, poses a real ethical question.
The founders say that their first customers were loved ones, aware that a human was in the loop.
Maybe. Maybe not.
This vagueness is indicative of a larger problem.
A mirror held up to the current AI ecosystem
Let's be honest. How many “AI” startups today:
- Are they selling technology that does not yet exist?
- Paste the word AI on a script, a manual process or a Google Sheet?
- Automate what they're selling today later?
Fireflies is no exception.
It is a Visible case of a widely used practice.
The major difference:
- Fireflies ended up building a real product,
- and this “fake” phase saved them from a seventh failure.
“Validation before automation”
Sam Udotong summarizes the approach himself :
“Validation before automation saved us from our 7th failure.”
This sentence is key. She says it all.
Fireflies first validated:
- that people wanted an automatic summary,
- that they accepted a third party in their meetings,
- that the perceived value was sufficient.
Technology came next.
A red line that is difficult to draw
This case raises a fundamental question for AI brands and startups:
👉 How far can we go before switching to AI-washing?
- Validating a need with humans in the loop is legitimate.
- Presenting this human as an autonomous AI is dangerous.
The border is thin. And the reputational risk is huge.
What Fireflies Really Teach Us
This case is neither a model to be glorified, nor a scandal to be blindly condemned.
Above all, it reveals that:
- AI is not a magic solution,
- value does not come from technology alone,
- usage often precedes automation.
But it also recalls an essential truth: Trust is a fragile asset.
A startup can survive an imperfect product. Much less at a loss of credibility.
Conclusion
Fireflies.ai became a unicorn despite its technologically misleading beginnings.
Not thanks to AI. Thanks to a detailed understanding of a real need.
This success raises a question that every AI player should ask themselves:
Am I selling a technological promise, or a use value?
The answer often determines the line between smart innovation... and dangerous drift.
.webp)
