Stop Separating Language Learning from Actual Use

01 /The Concept

Fluently built an AI assistant that improves spoken language skills for professionals who work in a language other than their native one.

The clever angle: it does this inside the workflow itself – specifically inside video calls, which have become an inescapable part of professional life.

Install the Fluently app on your computer and it starts monitoring what you say and how you say it during those calls. After each call, it delivers a detailed feedback report on your spoken performance.

No special integrations or bots required – the app runs locally and works with virtually any video conferencing platform. It also only captures and analyzes the user's own speech, not the voices of the other participants – so no one else needs to give consent.

Fluently saves a separate analysis for each call, letting users review individual sessions and track progress over time.

Every issue identified by the AI is linked to the specific moment in the transcribed conversation where it occurred – so feedback is always visible in context.

The AI doesn't just flag mispronunciations; it offers immediate practice exercises to correct them on the spot. It also tracks overuse of filler words and calls those out.

One of the most practical features for developing real conversational fluency: the app surfaces native-speaker synonyms that would naturally fit the context. Future versions will almost certainly expand this to phrasal verbs, idiomatic expressions, and colloquial patterns – the real markers of genuine language command.

Grammar correction isn't mentioned on the site yet, but it's almost certainly on the roadmap.

Each session also produces a summary proficiency rating using a standard scale – something like "C1 – Advanced" – giving users a clear snapshot of where they stand.

If no work calls are scheduled and a user wants to practice anyway, they can call the AI assistant directly and walk through whatever they want to rehearse: a startup pitch, a client presentation, a team update.

Individuals can use the app on their own, but the startup is putting particular effort into reaching companies that hire internationally and want to improve their teams' shared working language (currently English).

Fluently was founded last summer and made it into the current Y Combinator winter batch. The platform launched publicly a few days ago. Based on the website, the startup has raised not just the standard $500K from YC but additional capital from another fund.

02 /Why It Matters

There's no shortage of AI language learning apps. But what makes Fluently interesting – the real play here – isn't the AI-powered language coaching. It's that the learning is embedded into something the user is already doing.

A similar instinct drives Swap Language, [covered last fall](/review/tema-ne-tolko-bogataja-no-i-perspektivnaja): a platform for companies where employees from different countries learn the company's working language together. It raised €2M.

The broader problem with traditional education is that it trains people for a future that may never arrive. Two patterns account for most of the failure.

One is learning things with no immediate application – "I'll study French for that trip someday." The motivation is weak, there's no real practice context beyond rote exercises, and the effort fades.

The other is instructors cramming in as much content as possible because they equate quantity with quality. But knowledge without a chance to apply it stays airborne – theoretical. By the time an application moment arrives, the knowledge is gone.

The most important shift in modern learning: education embedded in practice:

- Learn what you can use right now.

- Learn what you're struggling with right now.

A key feature of this model is that knowledge transfers in small, immediately relevant pieces – "microlearning" in the pedagogical literature. But microlearning is a tool, not a philosophy. Delivering unnecessary information in small bites doesn't change the fundamental problem.

This applies well beyond language learning – to any professional skill.

A [review from last month](/review/bolshe-uchit-no-menshe-vkladyvatsja) covered Cloverleaf, which raised $15M for an "automated coaching" platform that sends employees small, timely nudges on communication, teamwork, and other workplace skills – timed to their actual calendar of upcoming meetings.

Last summer's [review of Sizzle](/review/ii-vzorvjot-obrazovanie-sovsem-s-drugoj-storony) covered an AI that helps high school and university students work through homework problems. It raised $7.5M in its first round. Sizzle's goal isn't to produce the answer fast – it's to teach while the student is doing the work, because doing homework is already a practical task that demands applying knowledge. If the knowledge isn't there, the right moment to acquire it is right now.

It brings to mind a thought experiment: what if universities reversed the model? Lectures and seminars would work through problem sets live. Independent reading of the underlying theory would be the take-home assignment. It might not be a worse model than the one we use.

03 /Opportunities

First takeaway: if you want an employee to develop a skill, teach it to them while they're actually doing the work that requires it.

And if you want people to push themselves to learn independently – give them a task that genuinely requires something they don't yet know. As a side benefit, you'll quickly learn who has growth potential and who has plateaued.

The direction for builders: platforms that embed learning into real activity.

Fluently embedded language development into regular work video calls. Sizzle embedded subject-matter learning into homework. What else could be learned, and where else could that learning live? That's a question whose answer could easily become your next platform idea.