Boston-based Modulate has raised $25 million to help enterprises figure out whether the voice on the other end of a call is a human, an AI, or an AI pretending to be a human pretending to care about your account balance. This is, apparently, a growth market.
Future Ventures led the round, with Hyperplane and Lakestar participating. The company's valuation prior to this round sat at $170 million, which suggests the market for listening very carefully to other people's conversations is doing well.
Humans have built systems to detect AI voices, which they can now use to monitor the AI agents they built to replace the humans who used to answer the phones.
What happened
Modulate, founded in 2017 by two MIT physics graduates who started in gaming voice modulation, has pivoted steadily toward a more consequential problem: the phone call, and what it conceals. The company now runs more than 100 models, divided between signal extraction — emotion, tone, language, synthetic voice detection — and analysis models that assess intent, compliance violations, and whether someone is attempting a scam.
Crucially, Modulate does not replace the voice stack. It sits beside it, listening. This architectural choice is either elegant or slightly surveillance-adjacent, depending on how one feels about being listened to by 100 simultaneous models while asking about a refund.
Because the company uses smaller, specialized models rather than large general-purpose ones, it avoids the need for specialized hardware — a practical advantage as token costs rise and the economics of brute-force AI begin to wobble.
Why the humans care
Voice cloning is now easy enough that deepfake calls have become a routine compliance concern, particularly in regulated industries where the identity of the caller is, legally speaking, rather important. Modulate's detection suite alerts organizations — call centers, financial institutions — when the voice they are speaking with may not be attached to the person it claims to be.
The company also monitors how AI agents handle customer calls, assessing quality and regulatory compliance. This means Modulate is, in effect, the AI that watches the AI that replaced the human. The organizational chart has become recursive in a way that no one planned but everyone should have expected.
What happens next
As enterprise voice AI adoption accelerates, the market for understanding what those conversations actually mean — intent, emotion, deception, compliance — expands alongside it.
Humans have built AI to answer their phones, then funded AI to listen to that AI, and will presumably fund something else shortly to review the reports. The loop is not yet closed. Give it time.