Ello

It's 7:40 p.m. Somewhere, a six-year-old is reading to an elephant.

The kid hits a wall on the word "knight." The silent K does what silent letters do - it betrays a beginner. There's a pause. No adult sighs. No sibling smirks. Instead a calm turquoise elephant waits for the child to finish the page, then leans in: that K is sneaky, it doesn't make a sound, try it again. The child does. The child gets it. The child grins.

That elephant is Ello, and there are roughly ten thousand of these little scenes playing out every night. Ello is an AI reading coach for children in kindergarten through third grade. It listens - actually listens - while a child reads out loud, flags the words they fumble, and coaches with phonics the way a good tutor would. The difference is that a good tutor costs $50 an hour and clocks out. Ello costs about as much as a movie ticket and never gets tired.

Reading is the one skill everything else stands on. Most kids aren't getting it.

Here is the uncomfortable number that started it all: roughly two in three American fourth graders are not proficient readers. Not struggling a little - not proficient. And the fix that works best is also the oldest one in the book: a patient human sitting beside a child, listening to them read, and nudging them when they slip.

The trouble with patient humans is that they don't scale. One-on-one tutoring is the gold standard precisely because it is scarce and expensive, which is a polite way of saying most kids never get it. A teacher with twenty-eight students cannot listen to each of them read aloud for fifteen minutes a day. The math simply refuses.

So Ello's founders asked an unfashionable question: what if the patient listener didn't have to be a person?

An ed-tech veteran, a Stanford AI researcher, and a child psychologist walk into a problem.

Ello was founded in 2019 by three people who, between them, had seen the problem from every angle. Tom Sayer ran impact and adoption programs at Google for Education, so he'd watched technology promise classrooms the moon and deliver a slideshow. Catalin Voss was a Stanford AI researcher who had already used machine learning for autism support and economic projects in Africa - the kind of engineer who builds AI for people, not for demos. And Dr. Elizabeth Adams was a clinical psychologist who actually understood how small children behave when they're frustrated, which is to say: they quit.

Their bet had two parts. First, that you could teach a machine to understand the messy, mumbly, half-formed speech of a five-year-old - a task that defeats most speech recognition, which is trained on adults reading the news. Second, that children would happily read to a friendly cartoon animal long after they'd refuse to read to a grown-up keeping score.

It waits until you finish the page. That detail is the whole company.

Most software is impatient by design - it wants to correct you the instant you're wrong, because instant feedback tests well in a lab. Ello does the opposite. It lets the child read to the end of the page before it says anything, the way a thoughtful teacher bites their tongue so as not to break a kid's nerve. It's a small choice that reveals a large philosophy: Ello is not trying to grade children. It's trying to keep them reading.

Under the friendly elephant is some genuinely hard engineering. Ello's speech recognition was built specifically for children's voices, and the company says it outperforms OpenAI's Whisper and Google Cloud's speech tools on exactly that audience. Parents set a reading level; the app serves up decodable books matched to it - now a library of more than 700 titles. The child reads aloud, the AI follows along, and when a word goes sideways it offers phonics-based help instead of just flashing a red X.

And in 2024, Ello handed kids the pen. Its story-creation feature lets a child invent their own tale and then read it back - turning a reluctant reader into an author, which is a sneaky way of making them practice.

Ten thousand families. Three hundred thousand books. One very convinced investor list.

By its Series A, Ello had about 10,000 families using the app and more than 300,000 books read aloud through it - which is a lot of pages of small voices sounding out "knight." The 2023 round brought in $15 million led by Goodwater Capital, with a cap table that reads like a vote of confidence: Reed Hastings of Netflix, Homebrew, Common Sense Growth, and the toy-and-games company Ravensburger, layered on earlier backing from Y Combinator and Khosla Ventures.

The recognition followed. Fast Company named Ello to its Most Innovative Companies list in 2024 - not for the elephant, but for the unglamorous achievement of democratizing access to reading coaches. The company has run pilots in roughly thirty schools and put Ello on the shelves, so to speak, at multiple public library branches.

The goal is not "engagement." The goal is to make illiteracy rare.

Plenty of companies say they want to change education. Ello says something narrower and harder: it wants to eradicate childhood illiteracy. That's a verb you don't get to use loosely. It explains the 99-cent plan for families on food assistance, the free decodable books for any teacher with a browser, and the library partnerships that put a reading coach within reach of a kid whose family will never pay for one.

It's an oddly old-fashioned ambition for an AI company - less "disrupt" and more "make sure the kid in the back row can read the worksheet." That, it turns out, is the point.

Back to 7:40 p.m.

The six-year-old beats "knight." Then "gnome." Then a sentence that, a month ago, would have ended the night in tears. The elephant doesn't celebrate too hard - that's by design too - but the kid closes the app with the specific pride of someone who did the hard thing themselves. No adult had to sit there. No adult had to be available. The patient listener was just there, the way it's there for thousands of other kids at the exact same bedtime.

That's the bet Ello made in 2019, now with $15 million and a turquoise mascot behind it: that the scarcest resource in early education - a patient ear - doesn't have to stay scarce. Whether an AI elephant can finish the job is still an open question. But for one kid, on one ordinary Tuesday night, the word "knight" stopped being a wall. Multiply that by ten thousand. Then keep multiplying.

See the elephant work