I was recently asked about which advancements in AI Audio excited me the most during the Tech.eu London 2025 conference last month.
We all get excited about the latest advancements and technological breakthroughs, but what exists today is already radically improving quality of life for thousands of people. More in this short clip:
I’m proud to be working at a company giving voices back to those who have lost them.
Full transcript:
Host:
James leads Creator Growth at the AI audio scaleup ElevenLabs and Itxaso, you’re a partner at Notion Capital investing in many AI startups and scaleups.
So James, we’ll start with you. Could you tell us a little bit about ElevenLabs’ journey? It’s raised so much money and scaled so quickly. I know you’re relatively new to the operation, but how have they managed to make that happen and swerve all the obstacles that so often come with scaling AI startups?
James:
ElevenLabs is an extremely fast-growing company. At its core is incredible voice technology — from day one, the focus has been on building some of the best AI audio models in the world.
The founders have also kept talent density incredibly high — small teams of very capable people working very hard to scale.
I’ve only been there about ten weeks, so I can’t speak much to what it was like two years ago, but really, it’s about great minds working very hard.
Host:
You’ve invested in so many AI startups and scaleups. Tell us about the scaling journeys you’ve seen across different verticals within AI.
Itxaso:
Yes, absolutely. I’ve been around for a little while — before Notion I worked at Microsoft doing investments in Europe. When I started in 2012, we used to ask companies: “Are you building on-premises or in the cloud?” Nobody asks that now. The same with “Are you mobile-friendly?”
That’s the transition happening with AI. It’s becoming everywhere — companies using data to better serve customers and build efficiency.
Personalization and efficiency are the two main things AI enables, and any company not doing that today is lagging behind. So yes, AI is in every vertical, and it’s very hard for us to invest in companies that aren’t leveraging data and AI.
Host:
That links nicely to my next question. Which AI innovations or breakthroughs excite you the most? James, I’m guessing your answer will be audio-related.
James:
Yes, definitely. We’re getting very close to real-time dubbing — where I could be speaking in English, and with almost no latency you’d hear me in French, German, or Mandarin. That’s incredibly exciting.
Similarly, we’re nearing real-time translation — imagine subtitles appearing instantly in any language.
But beyond that, our tech is already changing lives. People with ALS, for example, can clone their voices before they lose them — so they can still speak naturally. That’s moving and powerful.
Host:
That’s an amazing and particularly moving use case.
Itxaso:
I’m very product-led. I only invest in B2B companies and always look for tools that make employees’ lives easier and more productive — bringing consumer-like experiences to the workplace.
When I started my career as an engineer at Daimler in 2000, the systems we used were awful compared to today’s tech environments.
So I’m excited about tools that can exponentially improve how employees work and give them time back to focus on what matters.
Host:
When we talk about global, exponential growth — which ElevenLabs and many of your portfolio companies have achieved — what are the main barriers these companies face when scaling so quickly, and how do you help them tackle those?
Itxaso:
In AI especially, regulation is a big factor. Different countries have different frameworks, and that can be a barrier — but it can also be a benefit.
Take fintech: European regulatory frameworks allow “passporting” across countries, which helps expansion. The US doesn’t have that.
Regulation can slow you down, but it can also create a more trusted environment for adoption. For example, in AI, corporate clients need procurement and security checks — having shared standards actually helps startups sell faster.
Host:
How do you view Europe in this respect? Is it unique?
Itxaso:
Europe is cautious — but also pioneering. We led with GDPR and other privacy regulations. At first, these seemed like barriers, but they’ve also opened opportunities — especially in healthcare and finance.
Sometimes we overreact to regulation, but ultimately we adapt and make it work. Europe often leads the way in creating frameworks that later help collaboration and innovation.
Host:
James, when a company is scaling as fast as ElevenLabs, how do you balance R&D with business growth?
James:
We think of engineers in three groups at ElevenLabs:
- Research engineers – focused purely on improving the models.
- Product engineers – building the customer experiences.
- Growth engineers – working with me and the growth team to reach more users.
Interestingly, ElevenLabs doesn’t hire product managers. Growth engineers act almost like PMs themselves — they identify opportunities, run experiments, analyze data, and ship features independently.
For example, we discussed SEO performance in a meeting on Monday; one engineer proposed an idea, built it, ran an A/B test, and by Thursday we had 7% more signups. In a traditional company that might have needed four or five people.
We hire pragmatic, growth-oriented people who care about impact, not just code.
Host:
That’s fascinating. Itxaso, from an investor’s perspective, when you’re evaluating AI startups — especially in voice or general AI — what are your main criteria?
Itxaso:
First, I look for companies that own an entire workflow, not just a point solution. If a voice AI company solves four different voice applications, I want to see them integrate those — not rely on four separate providers.
Second, churn is a key metric. LLMs aren’t perfect, and many off-the-shelf models need fine-tuning. When users churn, that tells us the product isn’t delivering consistent value.
And third, the team. You need genuine AI experts who understand how to fine-tune, improve, and differentiate models. That’s essential for long-term defensibility.
Host:
Do you find it difficult to hire or back companies with that level of AI talent, given that big tech tends to snap them up?
Itxaso:
Not necessarily. The best talent doesn’t always want to work for big corporations. I worked at Microsoft — and that’s not the kind of place where you can propose, test, and launch something in a few days.
Startups give engineers direct impact. Once they experience that, they rarely go back to corporate life. So while big companies do attract some talent, innovative people often prefer the startup environment.
Host:
Can you share an example of a vertical company that’s nailed that end-to-end solution model?
Itxaso:
Sure — one of my latest investments is Nabu Nab, a French company like “TravelPerk for events.”
They built a tech platform connecting corporate procurement systems to a marketplace of venues, caterers, and suppliers.
It automates everything — budgeting, expense tracking, supplier approvals — with AI-assisted workflows. So employees can plan and manage events internally without external agencies.
It’s a perfect example of technology replacing services and driving massive efficiency.
Host:
That’s a great example — it really ties into efficiency and automation.
Moving from verticals to horizontals: James, how do you see AI audio impacting industries beyond tech — say, media or healthcare?
James:
That’s one of the most exciting parts. ElevenLabs is growing so fast because our tech applies across so many use cases — streaming, gaming, B2B, customer support.
Our first product was text-to-speech; now we’re doing conversational AI. Any company with a call center can use AI agents to triage or even resolve issues.
Audio is the most natural mode of communication — it was humanity’s first interface. Talking to your computer will soon feel more natural than typing or clicking. So I see AI audio becoming ubiquitous.
Host:
And Itxaso, from your broader view — how do you see AI evolving horizontally across industries?
Itxaso:
AI will be everywhere. The future is personalization through company-specific data.
There’s no reason, for example, why an email system shouldn’t automatically suggest the right folder or context when I open an attachment.
Right now, we’re still using off-the-shelf LLMs, but the opportunity is in building systems that learn from your behavior — tools that anticipate your needs.
I worked at Microsoft, and I dream of a world where Office 365 truly learns from my patterns and tells me on Monday morning exactly what I need to do. That’s the direction we’re heading in.
Host:
Thank you both — maybe we can check in next year at the summit in April and see which of these predictions came true.
This has been a wonderful conversation. Please give them a round of applause.