Sona8, a Y Combinator startup, turns what employees explain and show about their work into workflow maps that AI agents can use. Its interviewer asks follow-up questions and records how tasks run today, including exceptions. The company is built on a simple observation: the know-how needed to automate those tasks often sits in employees’ heads, email threads and old Excel files.

What Changed

AI-generated summary, reviewed by an editor. More on our AI guidelines.

How the interviews work

Interviews take 10 to 20 minutes, by phone or browser, in each employee’s language. Workers can explain a task, share their screen or point a phone camera at physical work. They can also let the agent record steps in the background while they work.

Most participants type their answers. Typed interviews now work like a chat, including on phones, with dictation available for anyone who prefers speaking.

The resulting map captures details such as how a quote gets approved and who gets called when a system is wrong. Teams can bring the maps into Claude, ChatGPT or their own agents through MCP, a connection for sharing information with AI tools. They use them to fix bottlenecks and to give those agents instructions.

Consulting customers

Sona8 is based in San Francisco and belongs to Y Combinator’s Fall 2026 batch. By its own count, it works with more than 20 organizations across Germany, Austria, the Nordics and the United States and has supported more than 10,000 employee interviews. Case teams at one of the world’s three largest strategy consultancies use it for stakeholder interviews and diagnostics.

In one deployment, the company says, participation reached 87.5%, and 400 interviews surfaced 35 AI use cases and 13 barriers.

Consulting projects are the starting point. Sona8 plans to expand to enterprises’ own strategy and transformation teams, keeping process knowledge inside the business after a consulting engagement ends.

Know someone who'd find this useful? ✉️ Email it to a friend in one click, or they can subscribe free here.

Founders and privacy

CEO Anton Hantel and chief product officer Thilo Tamme developed the idea on a BCG project in Hamburg. Consulting teams interviewed a few dozen people; they wondered whether an AI could interview everyone.

Hantel built Sona8 while pursuing an MIT MBA, then left when pilots needed his full attention. Tamme previously worked at BCG X and researched conversational AI at the Technical University of Munich. Chief technology officer Madeleine Malmsten was a principal engineer at McKinsey. The founders moved from Germany to San Francisco for YC.

German data-protection and employee consultation rules shaped the product. Tamme works with works councils and HR teams.

Screen and camera sharing stay off unless the customer enables them. The interviewer asks permission, and participants start each share and can stop it anytime. Shared material is stored privately in Frankfurt and does not appear in any screen, export or report.

Results are grouped by theme, with figures withheld for groups smaller than five. Employer reports leave out verbatim statements. Names, email addresses and phone numbers are removed before transcripts reach AI providers.

What follow-up questions find

Hantel and Tamme coauthored a field study pairing a survey with an AI voice interview on the same themes at a global consulting firm. In 18 of 44 first-session interviews, or 41%, a favorable survey rating was followed by a substantive concern about the same theme. Counting only the highest rating reduced that to nine interviews, or 20%.

Employees often softened concerns or limited how far they went. The conversation picked up concerns that the ratings alone did not capture. The study covered one firm and 44 first-session interviews, with no human-interviewer comparison.

Sona8’s longer-term plan puts the current map in front of an agent before it takes over a task, then updates the map as employees change how they work.

Frequently Asked Questions

What does Sona8 produce from employee interviews?

It creates maps of how work gets done, including exceptions and who gets involved when something goes wrong. Teams can bring those maps into Claude, ChatGPT or their own agents through MCP, a connection for sharing information with AI tools.

How do employees take part?

Interviews take 10 to 20 minutes by phone or browser, in each participant’s language. Employees can type, dictate or demonstrate tasks through screen and phone-camera sharing.

What can employers see?

Employers receive results grouped by theme, with figures withheld for groups smaller than five and verbatim employee statements left out. Shared screen and camera material is stored privately in Frankfurt and does not appear in screens, exports or reports.

Who founded Sona8?

Anton Hantel, formerly at BCG, is CEO; Thilo Tamme, formerly at BCG X, is chief product officer; and Madeleine Malmsten, previously a McKinsey principal engineer, is chief technology officer. The founders moved from Germany to San Francisco for Y Combinator.

What did the research find?

In 18 of 44 first-session interviews, employees gave a favorable survey rating and then raised a substantive concern on the same theme in an AI interview. Counting only the top rating, it was nine. The study covered one consulting firm and had no human-interviewer comparison.

AI-generated summary, reviewed by an editor. More on our AI guidelines.

OpenAI Cuts Voice AI Costs 20%, Adds Enterprise Features
OpenAI cuts voice AI prices 20% while adding enterprise features, but faces open-source rivals promising half the cost. The race for production-ready voice agents intensifies as integration complexity becomes the new battleground.
Andon Labs Tests AI Agents as Radio Operators
Andon Labs put Claude, GPT, Gemini, and Grok in charge of radio stations with money, tools, and listeners. The experiment showed model personality turning into playlists, sponsorship claims, and operational failure.
AI News

San Francisco

Editor-in-Chief and founder of Implicator.ai. Former ARD correspondent and senior broadcast journalist with 10+ years covering tech. Writes daily briefings on policy and market developments. Based in San Francisco. E-mail: editor@implicator.ai