Every AI tool you use is being sold to you cheaper than it actually costs to run and that’s a strategy. So who's really paying the difference, and what happens when they stop?

Recorded live at Latitude59 in Tallinn, Ailish sits down with Kart, a VC who's spent 12 years on the inside of those rooms. She co-founded her own firm, backed 30 funds and 15 companies, and now invests in early-stage deep tech. She also runs vibe coding workshops across Estonia, with particular success in women-only setups.

Kart pulls back the curtain on how the money actually moves and shapes the tech that shows up on our phones, laptops and everyday lives today. How a company goes from an idea with no product to something worth billions, why a valuation is basically a made-up number until the exit ("it's all trash until it's cash"), and what it really means when an AI company burns through $200 billion and investors keep writing cheques.

And we get into the bit that lands on all of us directly. Right now we're living in an era of subsidised tokens, where investors are footing the bill so you and I can use AI far cheaper than it actually costs. This won't last, so now is the moment to figure out where AI is genuinely useful to you, before the real price shows up. But what does that mean for how we use it now and how that might change things in the future?

In this episode, we get into:

• What a VC actually does all day, and how $500k on day zero turns into the apps on your phone
• Why every valuation is a bet on the future, and why half of companies don't make it
• The $200 billion question: are we in a bubble, or is this a bet on world domination that pays off
• Subsidised tokens, and the hidden cost behind the cheap AI we're all enjoying
• How AI has flipped the startup world, so distribution now beats technical skill
• Why vibe coding is opening the door for people who were shut out of building before
• The attention economy problem, and what happens now cold emails stop working
• Quick fire: the most overhyped buzzword, and who Kart thinks wins the AI race

Whether you're curious about where AI is heading, unsure who's really steering it, or you just want to understand the money shaping the tools you use every day, this one's for you.

0:01 — Live from Latitude 59, Tallinn: welcome to Trust Issues

0:34 — Meet Kart: 12 years inside the rooms that fund AI

2:17 — What a VC actually does, day to day

3:00 — The relay race: how money moves from idea to product

5:42 — What investors are really chasing

6:19 — Luck vs. skill: the Skype and Starship story

7:46 — How AI coding changed what VCs expect from founders

9:44 — When AI stopped being a sector and became the thing

12:35 — What a "valuation" actually means

17:17 — OpenAI's projected $200B burn

21:39 — The Uber playbook, and what it means for your AI bill

23:49 — Vibe coding as an unlock for women in tech

26:46 — Going public, explained

32:05 — The coming era of niche, AI-built software

39:10 — Why cold email stopped working

42:28 — Five years from now, if this goes well

44:52 — Quick fire round

ElevenLabs: Deepfakes, Darth Vader and How to Know If You Are Speaking to AI

AI can now clone a human voice so well that you might not realise you are talking to a machine. So how do you know who, or what, is really on the other end of the line? In this episode of Trust Issues, host Ailish McLaughlin sits down with Louise Meyer-Schönherr from ElevenLabs, one of the fastest growing AI voice companies in the world, to unpack how AI voice technology actually works, where you are already encountering it, and how to protect yourself from voice deepfakes and scams.

ElevenLabs builds the text-to-speech, speech-to-text and voice agent models behind a huge amount of the AI audio you hear every day. That includes dubbed YouTube videos, AI-narrated podcasts on Spotify, Darth Vader in Fortnite, and the voice assistants inside apps like Revolut and Klarna. Louise breaks down what voice cloning is, how ElevenLabs combines its voice models with large language models to build conversational AI agents, and the safety guardrails (watermarking, no-go voices, verification and moderation) designed to stop bad actors.

We also get into the big trust questions: the Joe Biden voice deepfake, whether AI voice is coming for voice actors' jobs, what happens to your data when you speak to an AI agent, and a simple trick for working out whether you are on a call with a human or an AI.

Whether you run a small business and are curious about AI receptionists, you are a creator thinking about audiobooks and dubbing, or you just want to stay discerning in an AI world, this one is for you.

In this episode:

• What ElevenLabs is and what voice AI actually does
• How AI voice cloning and text-to-speech work
• Where you are already hearing AI voices (Spotify, YouTube, Fortnite, Revolut, Klarna)
• Voice deepfakes, the Joe Biden calls, and how ElevenLabs blocks "no-go" voices
• How to tell if you are speaking to a human or an AI
• What happens to your data when you talk to a voice agent
• Are AI voices a threat to voice actors?
• ElevenLabs' mission to give one million people their voice back
• How small businesses can use AI voice agents for reception, bookings and support
• The future of voice as the primary interface for technology

Guest: Louise Meyer-Schönherr, ElevenLabs · Host: Ailish McLaughlin

Trust Issues is the podcast about AI, bias, trust and discernment.

You can't opt out of AI. It's already in your Uber, your Netflix, your news feed. So the real question is: do you want to understand it, or let someone else decide how it works for you?

Kate Minogue (ex-Meta, AI advisor, founder of The AI Leadership Lab) joins us to talk incentives, algorithms, user power, and why checking out of AI is the worst thing you can do right now.

SHOW NOTES

About the Guest

Kate Minogue is an AI advisor and fractional product leader with 6 years at Meta and a background spanning data science, gaming, fintech, and banking. She's passionate about helping non-technical business leaders get confident with AI, and recently launched The AI Leadership Lab, a course designed to do exactly that. Find Kate on LinkedIn or at kate-minogue.com.

In This Episode

• The Uber driver who checked out of AI (and why that's not actually possible)
• Netflix vs TikTok: same technology, completely different incentives
• Why understanding incentives is the key to trusting (or not trusting) AI
• AI hallucinations explained: what Kate told her sister that made her stop being scared
• How your data actually shapes the AI products being built
• Misinformation, deepfakes, and AI-generated content: which fears are warranted
• Why CEOs and graduates are behaving the same way around AI right now
• The "safe zones" framework for AI use in organisations
• How users (yes, you) can influence how AI develops
• US vs Europe: deregulation vs responsible AI as competitive advantage
• What teams actually want from leaders in the age of AI (it's not expertise)
• "Do it because the men are doing it and they are not apologising for it"

Mentioned in This Episode

• The AI Leadership Lab (Kate's course for non-technical business leaders)
• Max Tegmark (AI safety researcher, Web Summit talk)
• DeepSeek (Chinese AI lab)
• Sora (OpenAI's image/video generation app)
• EU AI Act and GDPR
• Boxer CEO memo ("AI is for you, not to you")
• Women in Africa building their own AI models (Web Summit)

Last week, Florence helped us get our heads around the right mindset for using AI. But there were a lot of words flying around. Agents. LLMs. Machine learning. What do those things actually mean? And more importantly, does it matter?

This week we're joined by Raji Ramakrishnan, a product leader at Lloyds Banking Group who works on agentic AI observability. Which, yes, is a mouthful. But by the end of this episode, you'll actually know what all of those words mean. And that's kind of the point.

Raji breaks down the entire AI landscape in a way that finally makes sense. She starts with the basics (AI is not magic, it's maths, data and programming) and walks us through how machines learn using an analogy that anyone who's taught a child flashcards will immediately get. Supervised learning? That's you holding up the flashcard. Unsupervised learning? That's the kid pointing at a cat in the street having figured it out on their own.

But this episode isn't just a glossary. It's about why understanding this stuff actually matters. Raji makes a compelling case that AI is coming whether you engage with it or not. Your mobile provider, your bank, your electricity company are all already using it. And the more you understand, the better equipped you are to know when to trust it and when to push back.

We also get into hallucinations (why AI confidently makes stuff up), the difference between generative AI and agentic AI, and what banks are actually doing behind the scenes to make sure AI doesn't go rogue. Spoiler: there are real humans watching.

In this episode, we cover:

1. AI, machine learning, deep learning, generative AI, agentic AI: what each one actually means and how they connect
2. The flashcard analogy: how machines learn in a similar way to children (supervised vs unsupervised learning)
3. Why AI is a prediction machine, not a truth machine, and why that distinction matters
4. Hallucinations: what they are, why they happen, and why you should always sense-check
5. Agentic AI: what changes when AI can take actions on its own, not just generate content
6. Observability and guardrails: what's actually happening inside banks to keep AI in check
7. Why jargon is an unnecessary barrier to entry and how to not let it hold you back
8. The mobile phone analogy: remember buying minutes for your Nokia 3310? AI adoption is on the same trajectory

This week we're kicking things off with a big question: is AI making us lazy? There's a study from MIT that suggests our brains might be outsourcing more than we realise. And with our brains not fully developing until around age 32, what does it mean that we're handing over so much cognitive work to AI tools before we've even finished cooking?

To help us figure it out, we're joined by Florence Jumpp, a product leader who's been working in AI and machine learning since 2019. Florence has a background in experimental psychology, and she's built her whole AI career around solving problems rather than obsessing over the tech itself.

Florence introduces us to her "drunk intern" framework. It's exactly what it sounds like. Think of AI as a capable but overconfident intern who's had a few too many. They'll absolutely get stuff done for you, but you wouldn't send them to the board meeting. And you definitely wouldn't have them work on your hardest problems.

She also shares her VEER framework for deciding which tasks to hand off to AI: looking at a task's Value, Enjoyment, Effort and Risk to decide whether it's a good one to hand off to AI.

In this episode, we cover:

1. Why thinking of AI as a "drunk intern" helps you use it more wisely (and why Florence's is called Jack)
2. The VEER framework for figuring out what to delegate to AI and what to protect
3. Cognitive offloading: why your brain has stopped taking notes in personal conversations too
4. How Florence uses Zapier to never face a post-holiday email wall again
5. Why doing the hard thing still matters, and how to force yourself to sit with the blank page
6. The positive feedback loop: using freed-up time to get even better at AI, not just filling it with more work
7. Why the people who think for themselves are the ones who'll stand out

About our guest: Florence Jumpp is a product leader specialising in AI and machine learning, with a background in experimental psychology. She brings a neuroscience lens to how we should think about AI's impact on our brains and our work.

Resources mentioned:

1. Zapier (zapier.com) for building AI-powered automations
2. MIT study on AI and cognitive offloading

A podcast for people who care just as much about how to use AI well as they do about what it means for us all.