Jordan Belfort’s new AI voice startup, Tone, wants to sell like the ‘Wolf of Wall Street’

Jordan Belfort, the man whose penny-stock brokerage Stratton Oakmont inspired "The Wolf of Wall Street," is now behind a voice AI company called Tone. He co-founded it with longtime investor Shervin Pishevar, according to an interview with TechCrunch. Belfort says the business has raised about $10 million in friends-and-family and seed money and has booked revenue of roughly $700,000, which he characterized as "probably about a million bucks." The company is also launching a self-service product aimed at content creators and authors who want to turn books into what it describes as expressive AI-generated narration, with voice cloning planned as a later feature.
Belfort's pitch rests less on voice quality than on delivery. He argues that in a phone conversation, body language collapses into tonality, and that roughly 15 distinct tonalities carry most of the persuasive weight in a sales call. Tone's selling point is deciding when to deploy them: phrasing a declarative statement as a question to build engagement, for instance, or dropping into a whisper to imply that information is scarce. Belfort says he began by teaching the model to reproduce his own voice and a single tonality, then added more and taught it to move between them. He calls data quality the crux, arguing that even a small amount of corrupt data undoes prior progress.
The system is built on small, lightweight in-house models rather than frontier ones, a choice Belfort ties directly to speed. He cites latency of about 600 milliseconds, which matters because a sales conversation stalls if the agent pauses noticeably before replying. The trade-off he acknowledges is intelligence: the small models, in his words, do not have quite as much of it. Tone's described workflow is end-to-end rather than assistive. In a health insurance sale, the agent asks qualifying questions about coverage type, budget and ages, selects a plan, walks through features and benefits, and responds to price objections by offering to lower the payment and explaining how raising a deductible would work. If the buyer dislikes the plan, it switches them to another.
Handling conversations that leave the script is the harder engineering problem. Belfort describes a "happy path" for each sale where every word can be mapped in advance, and says the real work is a model that can absorb interruptions, answer them well and steer back to the path without breaking rapport or fabricating claims, which in a sales context means lying. He is also pointed about tone quirks, saying the agent never opens with "that's a great question" unless it genuinely is, because the phrase reads as ingratiating. Tone's named customers are law firms and health insurance companies, with substantial use in polling and automobiles. Belfort frames the demand as capacity rather than replacement: clients have more leads than people to call them, or need coverage at hours and in languages their staff cannot provide.
On performance, Belfort is notably measured. He says Tone is not yet as good as a top-producing human salesperson, who would beat it call for call, but expects that to change within six months to a year, and contends it will significantly outperform a mediocre closer. That is a claim the company has not substantiated publicly, and neither founder is an engineer. Belfort names Ben Sussman as vice president of engineering and describes his own role as the arbiter of what sounds right, prototyping ideas that engineers then rebuild. He says the approach is patented, and positions Tone against ElevenLabs and Cartesia on the intelligence behind tonality choices rather than raw voice realism, predicting competitors will all sound human within a year.
Compliance shapes the product's boundaries. Belfort says outbound calling is limited to opted-in leads or existing customers, with polling cited as the exception where rules allow calls to private homes without opt-in. He points to the risk of litigation under telephone consumer protection law as the reason cold calling is off the table. For cloning, the planned safeguard is enrollment by speaking into the system, so the person creating the clone is known and signs consent; uploaded audio, which he discourages, would require matching a live voiceprint in real time. Pricing is mixed: per-minute for streaming, monthly subscriptions for the content platform, and outcome-based compensation for some clients where most of the fee depends on results.
Belfort's history is the obvious obstacle in enterprise sales. He pleaded guilty in 1999 to securities fraud and money laundering tied to Stratton Oakmont's pump-and-dump schemes, served 22 months in prison and was ordered to pay $110.4 million in restitution to 1,513 victims, completing that obligation in late April of this year, according to TMZ reporting cited in the article. Asked why a bank or hospital should trust the product, he says he made mistakes a long time ago and has spent more than 20 years teaching sales, and that clients unwilling to work with him are free to decline. Pishevar has faced misconduct allegations, which he denies. Belfort says he would entertain a substantial acquisition offer but plans to raise a significant round and does not intend to sell soon.
Why it matters: Tone lands in a crowded voice AI field, and its differentiation is application logic, such as which tonality to use and when, rather than voice quality, which is where the competitive edge is hardest to prove and easiest to copy. Developers building conversational agents should note the latency-versus-capability trade-off Belfort describes, since sub-second responses on small models come at the cost of reasoning depth. For product teams, consent handling for cloning and strict opt-in discipline for outbound calls look like table stakes rather than features. Whether enterprise buyers separate the technology from its founders' records is an open question, and my inference is that regulated sectors such as insurance and health care will test that separation hardest.
Based on reporting from the original publisher. Visit the source for full context and later updates.
Publisher excerpt
Belfort claims his AI sales voice can already close deals almost as well as the best human salespeople, and he thinks it will catch up in a year.