Nagish.

Somewhere right now, a man is on hold with his pharmacy. He cannot hear the hold music, and he doesn't miss it. When the pharmacist picks up, the words appear on his screen as text, a half-second behind the voice. He types his reply; the pharmacist hears it spoken. Neither of them thinks about it much. That ordinariness is the whole point of Nagish.

Nagish is a New York company that does one stubborn thing well: it lets people who are deaf or hard of hearing make and take phone calls by themselves. The app sits on a phone you already own, links to a number you already have, and turns speech into text and text back into speech in real time. The caller on the other end installs nothing and, often, notices nothing. For roughly 48 million Americans with some degree of hearing loss, that is the difference between calling the doctor and asking someone else to do it.

A $2 billion workaround nobody loved

For decades, the way a deaf person made a phone call was to invite a stranger into it. A human relay operator - a captioner, a stenographer, sometimes a sign-language interpreter on video - sat in the middle, listening to one side and typing or signing it to the other. It worked. It was also slow, and it meant a third party heard your call to the bank, the clinic, the lawyer. The U.S. government funded this through a relay pool that ran into the billions of dollars a year, paying humans to do what was, at its heart, transcription.

The founders' insight was unglamorous and correct: the operator was never the feature. Privacy was the feature, and the operator was the tax you paid to get a call at all. Remove the human, keep the accuracy, and you don't just cut a cost - you hand the conversation back to the two people who were supposed to be having it.

Caption: The relay industry spent years perfecting the middleman. Nagish spent its years perfecting his absence.

Two engineers, one inconvenient question

The idea started in 2019 as a thought experiment. Tomer Aharoni, then a computer science student at Columbia University, found himself wondering how someone who cannot hear or speak actually makes a phone call. He asked his friend and fellow engineer Alon Ezer. The answer they uncovered was blunt: mostly, deaf people just didn't make phone calls. The technology to fix it - fast, accurate, on-device speech recognition - had only recently become good enough to trust with a live conversation.

They incorporated Nagish in 2021. The name is Hebrew for "accessible," which is either a mission statement or a spoiler, depending on how you read it. Both founders were engineers, so they did the engineer thing and built the relay service themselves rather than reselling someone else's.

What you can actually do with it

Open the app and your calls arrive as readable text, line by line, as the other person speaks. Type back and Nagish speaks your words aloud to them in a natural voice. You keep your existing phone number through number linking, so people reach you the way they always have. A personal dictionary teaches it the names and jargon that generic speech engines mangle. Transcripts are saved locally on your device, a built-in spam filter screens the junk, and Bluetooth devices are supported for people who use them.

Crucially, there is no operator on the line and no relay service eavesdropping. Nagish describes it as the first end-to-end private relay service - the conversation stays between caller and recipient. For people who'd resigned themselves to handing their private calls to a stranger, that's less a feature list than a small restoration of dignity.

Caption: Four numbers, one argument - the market is enormous, the price to the user is zero, and nobody else is listening.

Money, regulators, and the founders of Datadog

In July 2024, Nagish announced $16 million in total funding. The headline was an $11 million Series A led by Canaan Partners, layered on top of an earlier $5 million seed. The cap table reads like a vote of confidence from people who build infrastructure for a living: alongside K5 Global, Tokyo Black, Cardumen Capital, Vertex Ventures Israel, Contour Venture Partners and Precursor Ventures, the round drew the founders of Datadog and Looker - two companies that know something about turning messy real-time data into something useful.

The other proof point is regulatory, and it matters more than it sounds. Nagish is certified by the Federal Communications Commission to provide telecommunication relay services. That certification is why the consumer app can be free: like other relay providers, Nagish is reimbursed from the federal relay fund. Few companies hold it. It is the kind of unglamorous moat that takes years and is hard to copy.

The business splits cleanly in two, which is part of why investors liked it. On the consumer side the app is free and the federal reimbursement does the paying, so growth costs the user nothing and the company gets compensated per call rather than per subscription. On the enterprise side, Nagish sells a fuller suite of accessibility and transcription tools to businesses and charges based on usage. One side is mission; the other is margin. The two happen to be the same technology, which is the sort of tidy arrangement that rarely survives contact with reality but, here, mostly does.

Who actually uses it tells the same story from the ground. The core audience is the wide and underserved population of deaf and hard-of-hearing Americans - not a niche so much as a market everyone else quietly wrote off. Around it sits a team of roughly 43 people, remote-friendly and engineering-led, the kind of company small enough to ship fast and serious enough to clear an FCC audit. Competitors exist - InnoCaption, CaptionCall, ClearCaptions, Hamilton CapTel and the older human-relay services - but most were built around the operator Nagish set out to remove.

Caption: Venture math is usually dull. Here it underwrites a simpler claim - someone with money believes deaf people deserve a private phone call.

"We make communication more accessible"

The company's stated mission is plain enough to fit on the homepage: make communication more accessible, and let people with hearing loss communicate independently and privately. Independence is the operative word. The point was never to give people a better operator. It was to remove the need for one, so a phone call stops being a logistics problem and goes back to being a phone call.

There is an obvious next chapter, and Nagish has started writing it. Nagish Live Transcribe extends the same engine beyond the phone - free real-time captioning for the world that happens out loud in front of you, not just down the line. The phone was the hardest problem to solve first. Everything spoken is the larger one.

The conversation, returned

Skeptics will note the obvious: speech recognition still trips on accents, crosstalk, and bad connections, and an AI relay is only as good as its worst transcription. Fair. But the trend line is the company's friend - the models get better while the humans they replaced never could scale. And the regulatory blessing means Nagish isn't betting on charity or goodwill; it's plugged into the same fund that paid the operators it's retiring.

So return to the man on hold with his pharmacy. A few years ago he'd have needed an operator listening in, or a hearing relative to make the call for him, or he'd simply have skipped it and driven over. Today the words just appear, he types back, and the pharmacist hears a voice. He hangs up without thinking about any of it. Nagish has spent $16 million and several years of two engineers' lives to make a phone call boring again. For 48 million people, boring is the most radical thing it could possibly be.