Why this exists
Your company still runs on people passing messages
I've seen companies that don't. A handful of people do the work of a much bigger team, and everything the company knows sits in one place that every person and every agent can read. Ambient exists to bring that to every company.
What I saw in San Francisco
A while ago I spent time in San Francisco at Founders Inc., one of the city's best accelerators. I met more founders than I can count, and I got a close look at how the best startups in Silicon Valley and the big AI labs actually work.
The best teams didn't look like any company I knew. A handful of people shipped what used to take a whole department. Their meetings were recorded and transcribed. Agents read everything, did the recurring work overnight and asked a person only when they had to. Nobody wrote status updates, because nobody needed them.
They called it AI-native. It was the clearest picture I'd seen of where every company is going.
Most companies still run like a Roman legion
The org chart is older than you think. A Roman legion put eight soldiers to a tent, ten tents to a century, centuries into cohorts and cohorts into a legion. Orders went down the chain, reports came back up, and at every step a person carried the message.
Two thousand years later, most companies work the same way. Meetings move information from one head to another. Managers spend their days routing it. People are the wiring.
Take a normal Tuesday. At 9:30 your team moves the launch to Friday. By lunch, the three people who weren't in the room are still planning for Tuesday. At two, someone asks Claude for the customer update, and Claude has never heard of the launch. At five, a new teammate asks why the price changed. The one person who knows is on holiday.
Nobody did anything wrong. This is the human relay, and it's how almost every company works.
Why it hurts more now
The relay worked while people were the only ones who could read, remember and decide. AI changed that, and most companies bolted it onto the old structure.
A chat window for questions. An agent that works until it gets stuck, then waits for a person. Every step waits for someone, so the work stops at night and starts again in the morning. And your best people spend their days pasting context into Claude and pasting its answers into Slack. The software engineer Niklas Gruhn found the name for the person in the middle, and it stuck:
“I can talk to Claude myself. It's going to be faster and I get to control the context. I don't need a meat proxy in between.”
Nobody chooses to be a meat proxy. The company makes them one, because every agent knows only what somebody pasted in.
The AI-native company
An AI-native company is built around a brain that every person and every agent reads. Loops run on top of it and do the recurring work. People work at the edge.
It rests on five building blocks.
Everything is legible. Meetings are recorded and transcribed, with everyone's consent. Decisions are written down with their reasons. Work happens where AI can read it. If it isn't written down, for AI it didn't happen.
One brain. All of it flows into one place that stays current: what the company decided, who owns what, and how it handles the edge cases no procedure ever captured. Every person and every agent reads from it.
Loops do the recurring work. A loop takes signals from the real world, like support tickets, usage, billing and code changes. It acts within clear rules, uses the company's tools, has its work checked by a second model and measures what its change did. Then it runs again, so the company improves while the team sleeps.
People work at the edge. People do what only people can: talk to customers, make the hard and new calls, set the direction and earn trust. Nobody's job is to carry information. It reaches whoever needs it.
Small teams with clear owners. There are no layers of managers routing updates. Builders bring working prototypes to meetings, and one person owns each outcome. The company spends tokens on loops before it adds headcount.
What sets it apart
- A traditional company
- What the company knows lives in people's heads, and the org chart routes it. Work waits for whoever knows.
- An AI-assisted company
- The same structure, with a copilot for everyone. People get a little faster, and the agents still start from zero.
- An AI-native company
- What the company knows lives in its brain. Loops do the recurring work, people work at the edge, and the company gets a little better every night.
I rebuilt Miora this way
Back home, in the first half of 2026, I turned Miora, where I'm co-founder and CTO, into an AI-native company. Our meetings were transcribed into a company wiki. Agents read it and kept it current every night. They wrote our reports, checked our work and asked us when they needed us.
It changed how we worked more than any new model did.
It also took weeks of engineering, and it needs an engineer to keep it alive. Once, one of the nightly jobs stopped quietly, and nobody noticed for seven weeks.
Why most companies can't get there
A new startup can build this way from its first day. A company that already runs can't. It has customers, deadlines and a team that has worked a certain way for years. It can't stop to rebuild itself, and it can't hire an engineer just to keep a brain alive.
That's most companies. They're who we're building for.
What we believe
Every agent should know what your team knows. An agent is only as good as what it knows about your company. Every agent should know what the others, people and agents, did and decided.
A company's memory should write itself. Nobody should have to take notes, file them or keep a wiki current. The work your team already does should be enough.
People belong at the edge. Carrying information is work for software. People should spend their time with customers, on hard decisions and on where the company goes next.
AI is multiplayer. Today everyone works alone with their own agent. The teams that pull ahead will be the ones whose agents know what everyone else's did.
What your company knows belongs to you. It's plain Markdown in your own repository, readable without us and yours to take anywhere.
What we're building
Ambient gives a company a brain without rebuilding the company. It records your meetings on your iPhone and Mac and captures the work your team does with AI agents. It turns all of it into one wiki that stays current on its own and lives in your own repository, and Claude, ChatGPT and Copilot read it over MCP. Routines run on top of it on a schedule, and we build agentic loops for your team's recurring work.
Recordings are transcribed on your iPhone or Mac, and the audio never leaves the device. Everything we store stays in the EU.
Where this goes
In a few years, every company will be AI-native, or competing with one that is. The question is how they get there.
We want a team of ten in Munich, in Vienna or in a valley in the mountains to work like the best teams I met in San Francisco. Nobody asks “did that ship?” anymore. No agent starts from zero. And people spend their days on the work only people can do.
Nikolai Madlener