Mati Staniszewski: The Founder Playbook

Mati Staniszewski is the Co-founder and CEO of ElevenLabs, the AI audio company behind human-like text-to-speech and voice agents, and founders study him because he took a domain most researchers ignored and turned it into an $11 billion business in under four years. If you want to understand how a small team reaches the frontier of a category, mati staniszewski is one of the clearest cases to model, which is why he sits alongside the other tech founders in this roster.

The facts set the stakes. Staniszewski, a Polish entrepreneur born in 1995, holds a BSc in Mathematics from Imperial College London and spent his early career at Opera Software, BlackRock (helping launch the Aladdin Wealth platform), and Palantir (as a Deployment Strategist from 2018 to 2022). In 2022 he left to co-found ElevenLabs with his high school friend Piotr Dabkowski, now CTO. By February 2026 the company had raised a $500M Series D led by Sequoia at an $11 billion valuation, crossed $500M ARR, and put Staniszewski on the TIME 100 Most Influential People in AI. Here is what a founder building with AI can take from how he did it.

Pick the overlooked frontier, not the crowded one

When ElevenLabs started, audio was treated as a niche. It was a large domain, but few researchers were working on it seriously, which is exactly what made it a good place for a small startup to plant a flag and become the frontier lab for the entire category. Everyone else was fighting over text and image; the space with the best odds was the one nobody was crowding.

The wedge was personal. In ElevenLabs' Mati Staniszewski: How Voice Becomes the Interface for AI, Mati says "The first moment for us... was when we could replicate my voice based on a good sample, that was like a first wow moment to to myself." That single demo, cloning his own voice from a sample, told him the technology was real and the category was open. He also frames the mission plainly: in the same talk he says "We believe there's a little bit of that to really fix text to speech and fix that emotionality." The problem was known and unsolved, and that was the point.

So what: do not pick the topic with the most papers and the most competitors. Pick the large domain that is thin on serious builders, prove one undeniable demo to yourself first, and you can own a frontier a bigger company will not fight you for early.

Keep the team small and flat so decisions ship

ElevenLabs reached an $11 billion valuation and $500M ARR with a team measured in the hundreds, not thousands. The company deliberately keeps its groups small and flat, which lets decisions get made and shipped quickly instead of dying in layers of approval. Agility is treated as a structural choice, not a personality trait.

That structure is a big reason the product pace looks the way it does. Staniszewski's own read of the company's odds is grounded in the people, not the headcount: in ElevenLabs' Mati Staniszewski: How Voice Becomes the Interface for AI he says "I have the most luck in the story of 11 Labs because well, it started in 2022 it felt feels like it started 17 years ago when I met my my co-founder Piotr." A tight founding relationship and a flat team let the company move at a speed larger orgs cannot match.

So what: resist the urge to grow headcount as a proxy for progress. A small, flat team where the person doing the work also makes the call will out-ship a larger org that routes every decision through a chain. Design the org chart for speed before you design it for scale. If you are a lean team, see solo founders who build like a team for how to punch above your size.

Hire for the work, wherever the person is

ElevenLabs does not hire by zip code. The company finds the best researchers based on what they have actually built and brings them on wherever they happen to be. In a field where a handful of people move the frontier, the constraint is talent density, not proximity to an office, so the hiring filter is the work itself.

So what: write your hiring bar around demonstrated output, a paper, a model, a shipped project, and then remove location from the equation. In a thin, specialized domain, the person who has already done the hardest version of the work is worth more than ten generalists near your office. Model the filter on the work, not the resume line or the commute.

Embed engineers in every function and automate the workflow

ElevenLabs pushes engineering into the non-technical parts of the company. Integrating engineers directly into non-technical teams has improved efficiency and innovation across the organization, because the person who feels a manual, repetitive workflow is sitting next to the person who can automate it. The result is that internal processes stop being human bottlenecks.

Mati describes the payoff directly. In ElevenLabs' Mati Staniszewski: How Voice Becomes the Interface for AI he says "We just made it so much easier and of course that's fully automated now with how we work across that team." The pattern is to embed the engineer, find the friction, and automate the workflow until it runs itself.

So what: do not silo your engineers in a product org and leave sales, ops, and support to grind through manual work. Put a builder inside each function, let them automate the repeated steps, and you convert headcount you would have hired into leverage you already have.

Hold voice to a human bar before calling it done

Staniszewski's product thesis is that voice becomes the default interface for AI only when it clears an emotional bar, not just an accuracy bar. Agents have to match the user's emotional state and earn trust before people will talk to software the way they talk to a person. Getting the words right is table stakes; getting the emotion right is the actual product.

That is why the strategy is a model plus an ecosystem, not a model alone. In ElevenLabs' Mati Staniszewski: How Voice Becomes the Interface for AI he says "We try to train the models, build a product and an ecosystem that will drive of course the biggest impact for all our customers, all users." The frontier model is the engine, but the product and the ecosystem around it are what turn a wow demo into something people rely on.

So what: pick the quality dimension your category actually judges you on, for voice it is emotionality and trust, and refuse to ship until you clear it. Then wrap the model in a product and ecosystem so the capability reaches real users. If you are building anything conversational, our guide to AI voice agents and the field notes in AI cold calling show what that human bar looks like in practice.

What to copy this week

• Audit your market for the large domain that is thin on serious builders, and aim there instead of the crowded frontier everyone is already fighting over.
• Prove one undeniable demo to yourself before you scale anything; if it does not make you say wow, keep working on the wedge.
• Keep your team small and flat, and make sure the person doing the work is the person who makes the call, so decisions ship instead of stalling.
• Rewrite your hiring bar around demonstrated work and drop location as a requirement, then embed an engineer inside a non-technical function and automate its most repetitive workflow.
• Define the one quality bar your category judges you on (for voice, emotion and trust) and refuse to ship until you clear it.

AI Operating System for Startups

Sources

• ElevenLabs' Mati Staniszewski: How Voice Becomes the Interface for AI
• Wikipedia: Mati Staniszewski
• TIME 100 Most Influential People in AI 2025: Mati Staniszewski
• TechCrunch: ElevenLabs raises $500M from Sequoia at an $11 billion valuation
• ElevenLabs blog: Series D announcement
• Mati Staniszewski on LinkedIn