Podcast Transcript: Lightpanda, AI Agents & the Future of Automation
Host: Welcome everyone to a new episode of “Y'a pas que la data” (There’s More Than Just Data). Today, we have the pleasure of welcoming Katie Hallett, co-founder and CEO of Lightpanda. Lightpanda is a Paris-based startup working in the AI space, focused on web automation, headless browsers, and agents.
Katie, welcome to the show!
Katie: Thanks for having me. Happy to be here!
Host: Could you start by telling us a little bit about yourself and your background?
Katie: Of course! I’m originally from the UK, but I’ve been living in Paris for over ten years. My background is mostly in SaaS and web-based businesses. Before Lightpanda, my co-founders and I built another startup together, which was more of a traditional SaaS e-commerce business. But last year, we decided to start something new—Lightpanda.
What Is Lightpanda?
Host: Can you explain what Lightpanda is and why you started this project? Because I have a lot of questions, but let’s start from the basics.
Katie: Sure! Lightpanda is a headless browser specifically designed for machine use. Right now, the main solution for web automation is Chrome in its headless mode. The key difference between Chrome Headless and Lightpanda is that we’ve completely removed graphical rendering. That means we don’t just get rid of the user interface, but also the entire rendering engine. As a result, Lightpanda is much lighter and significantly faster for machine-based tasks.
Host: Is this a well-known problem, or is this a niche solution? Because I had never heard of this before Lightpanda.
Katie: It’s a known problem, but primarily among developers working on web automation. In our previous company, we scraped millions of web pages daily and often used Chrome Headless for that. Many people in web scraping and automation know this issue well. The modern web is full of JavaScript, and traditional scraping techniques no longer work. That’s why a headless browser is necessary—to execute JavaScript and retrieve content just like a human would.
The Challenges of Web Scraping & Automation
Host: For our listeners who might not be familiar with web scraping, can you explain why this is such a big challenge?
Katie: Sure. The web today is very different from what it was 10–15 years ago. Back then, most pages were static HTML, and you could simply extract data with simple scripts. But now, JavaScript is everywhere, dynamically generating content. If you don’t use a real browser to execute the JavaScript, you miss a huge portion of the web’s data.
Host: So that’s why companies need headless browsers?
Katie: Exactly. Many businesses rely on automated web data collection for competitive intelligence, price monitoring, search engine indexing, and more. Google, for example, has its crawlers for indexing pages, but there are many other use cases where companies need to interact with web content at scale.
Host: But isn’t there already an entire industry built around web automation? We’ve had Selenium, Puppeteer, Playwright… Why build a new browser?
Katie: Good question! Most web automation tools today still rely on Chrome or other full-featured browsers. These are designed for humans, not machines. Even Chrome Headless still includes a rendering engine that is unnecessary for machine-based automation. That extra complexity makes it slow and resource-intensive. Lightpanda is purpose-built for automation, which is why it’s 10x faster and 10x lighter than Chrome Headless.
The Growing Role of AI Agents
Host: That brings us to AI agents. Your website mentions three key promises: large-scale web scraping, automation, and AI-powered agents. Is Lightpanda really necessary for AI agents?
Katie: That’s a great question. First, we need to define what an agent is. Right now, people throw around the term a lot without a clear definition. To me, an agent is an AI system capable of acting autonomously beyond predefined rules. It can reason and make decisions on behalf of a human.
Host: So where does Lightpanda fit into that?
Katie: Most AI agent demos today use full browsers like Chrome, which is incredibly inefficient. For example, Operator from OpenAI runs a GUI-based Chrome instance for every action, which is slow and resource-heavy. If AI agents are going to scale, they’ll need much lighter execution environments. That’s where Lightpanda comes in—it enables automation at scale without the performance overhead of traditional browsers.
The Economics of AI Agents
Host: Let’s talk about business models. How do you think AI agents will be monetized?
Katie: I see two main sides to this: how users will pay for AI agents and how companies developing them will make money.
For users: AI agent pricing will likely be value-based, meaning you’ll pay per completed action rather than a subscription fee.
For companies: Monetization could come from transaction fees (taking a percentage of financial transactions completed by agents) or from sponsored results, where businesses pay to have their services recommended by AI agents. But this raises ethical concerns around bias and neutrality.*
Open Source & The Future of AI
Host: Speaking of monetization, Lightpanda is open-source. Our last guest, Hervé Bredin, asked whether open-source AI is beneficial or risky. What’s your take?
Katie: It’s a fascinating question, especially given the recent debates around AI openness. Sam Altman himself admitted that OpenAI is “on the wrong side of history” when it comes to open source. Personally, I think open-source AI is not just beneficial—it’s necessary. Open-source software is often more secure because vulnerabilities can be identified and fixed faster. Keeping AI closed-source doesn’t prevent bad actors from misusing it; it just limits oversight and accountability.
The Future of the Web & Human Interaction
Host: One last philosophical question—if the web is increasingly built for machines rather than humans, does that mean we’re losing creativity? Websites today are visually engaging, but if AI agents are the primary consumers, does that push us toward a more uniform, machine-readable internet?
Katie: That’s a great point. But I think it’s actually a reflection of how much the internet has shaped human interaction. If AI can handle repetitive online tasks, maybe that frees humans to engage more meaningfully with the real world. The web was originally designed to mimic the physical world, with concepts like “storefronts” and “addresses.” Now, AI is pushing it toward a new paradigm—one optimized for machine interaction. That doesn’t mean creativity disappears; it just shifts to new mediums like augmented reality and spatial computing.
Final Question for the Next Guest
Host: We always end with a question for our next guest. What’s yours?
Katie: I love controversial topics, so here’s mine: Inspired by the recent news around Project Stargate, how do we measure the ROI of AI? We’ve seen massive investments—$500 billion for Stargate, for example—but how do we determine if it’s truly worth it? We’ve talked about AI accelerating drug discovery and vaccine development, but is that the real benchmark? What should success look like?
Host: That’s a fantastic question. We’ll make sure the next guest tackles it. Katie, where can people follow your work?
Katie: You can check out Lightpanda on GitHub, follow our updates on LinkedIn, or reach out to me directly.
Host: Perfect. Thanks again for joining us, and see you all in the next episode!