Conversation analytics spent years answering a narrow question: what did the customer say? Modulate has raised $25M to argue the more useful question is why they said it, and whether a human said it at all.
Future Ventures led the round, with Hyperplane and Lakestar joining. PitchBook data puts the Boston company’s earlier fundraising at $41M and its prior valuation at $170M.
The company dates to 2017, when Mike Pappas and Carter Huffman started it after meeting as physics students at MIT.
Voice modulation for games was the first product. Voice-based moderation followed. Understanding a conversation, rather than transcribing it, came last.
The platform runs more than 100 small models split into two jobs. One group extracts signal, reading emotion, tone and language, and judging whether a voice is synthetic. The other assesses intent, flagging scams, rule violations and what a caller actually wants.
That pairing puts Modulate beside the voice stack enterprises already run. Customers include call centres that want warning of deepfake attempts and businesses that need to know whether an AI agent handled a customer well and stayed inside compliance rules.
Huffman says the small-model approach avoids specialised hardware and heavy compute, and makes it cheap to add new capability as it arrives. He also argues emotion analysis must go past surface readings, since customers often stay polite while being deeply dissatisfied.
The company employs 40 to 45 people and plans to add about ten, with on-premises and on-device deployment in development. It also markets the technology for spotting cyberattacks that arrive by phone.
