All my experience
My startup · 2023 to 2024

VoiceKittTalk to your car like Hasselhoff talks to KITT.

A generative AI voice assistant for business software. We started with field sales reps, people who spend hours a day in the car, so they could get their work done just by talking.

Aemal Sayer and Alexander Blum, the VoiceKitt co-founders, in the KITT car
Alex and me in KITT
  • FoundedJanuary 2023
  • Based inBerlin, Germany
  • My roleCo-Founder & CEO
  • AcceleratorTechstars Berlin '23 with Audi
The story

I have always been a slow reader. Much later in life I found out that I am probably dyslexic, and in August 2026 I was diagnosed with ADHD. So voice was always my way in. For years I built voice dictation tools for myself, and voice ended up in almost everything I made.

24Geeks was a voice company early in its time: if you were stuck on a bug, you could start a live voice call with a senior developer right in the browser. I built it alone on WebRTC, a full Google Meet-like experience with recording, before Google Meet became famous. From 2019 to 2023 I built zooc, an asynchronous voice platform that helped teams communicate more efficiently.

I also spent years in automation: robotic process automation, browser automation, little Chrome extensions that did the boring clicks for me. The question I kept asking was: once I can transcribe what someone says, can I understand what they mean? That question goes back to my master's in Web Science, where our very first assignment was to analyse the text of a tiny Wikipedia.

Then ChatGPT arrived. Large language models could finally understand intent, and I already knew voice. So in January 2023, very naturally, I started building VoiceKitt with my co-founder Alexander. The name comes from KITT in Knight Rider: talk to your car like David Hasselhoff talks to KITT.

We started with field sales reps, who drive around four hours a day and still end up with hours of admin in the evening. You say “reply to Peter's email from this morning and attach the price list”, and VoiceKitt drafts the reply, finds the file and sends it when you confirm. It worked with Salesforce and Microsoft 365, in German and English.

We built it at a time when there were hardly any AI agent tools. Our first versions took 13 seconds to answer, and we brought that down to around 500 milliseconds ourselves. We went through Techstars Berlin with Audi Denkwerkstatt and won our first paid pilot customer, Merck.

Looking back, I almost unknowingly built the core of something much bigger. If we had gone horizontal and offered our voice layer as an API, instead of going vertical on Salesforce and business software, we could have become a Vapi or an ElevenLabs.

In the end the funding risk was too high, and we had to liquidate the company and let it go. I lost around $500,000 on VoiceKitt. Most of it came from my personal savings and the profits of my previous companies, and $120,000 came from the Techstars accelerator.

voicekitt.com
Aemal Sayer, Co-Founder and CEO, and Alexander Blum, Co-Founder and COO, on the VoiceKitt team slide
Watch it work

Demos and pitches.

Product demo: talk to your car like David Hasselhoff talks to KITT

  • Techstars '23 Demo Day pitch

  • Pitch at Chatbot Summit 2023 in London

  • Pitch at IdeaLab 2023 at WHU Vallendar

Behind the scenes

The app, the deck and the car.

Want me to train your team?

Today I teach teams and founders how to build AI-native companies with AI employees, in hands-on workshops and team trainings.