What do you do when millions of viewers are waiting, and the stream fails?
For Netflix chief technology officer Elizabeth Stone, moments like that reveal how teams respond under pressure, take ownership, and keep moving forward.
At Elevate Talks during Toronto Tech Week, she pulled back the curtain on how Netflix navigates AI, live streaming, and risk, and shared lessons for any leader facing digital transformation.
Stone joined Netflix in 2020 and was appointed CTO in 2023. Now she leads a wide portfolio including engineering, data science, consumer insights, and product innovation. Her conversation highlighted the complexity of expanding a platform while maintaining an intuitive member experience.
“Hopefully for members at Netflix, it doesn’t seem as complicated as it actually is,” said Stone. “It takes a lot of work to make it run that smoothly.”
That simplicity, she explained, is built on internal principles of curiosity, iteration and long-term thinking.
Stone pointed to Netflix’s experimentation with natural language search (for instance, allowing users to type phrases like “funny, but not too funny”) as an example of designing AI with empathy. For business leaders, it’s a reminder that the most powerful data products often start with listening, not just optimizing.
These features, Stone noted, come directly from ongoing research with members about what helps them feel more connected to the content they discover.
From missteps to momentum: The live streaming learning curve
When Netflix attempted to stream the Love Is Blind reunion live in 2023 and it failed to load for many viewers, the company received immediate and widespread criticism. But inside Netflix, the setback was treated as a lesson.
“I would much rather be at a company that takes that type of risk and innovation than one that doesn’t,” said Stone. “Otherwise, we’d only take the safety bets.”
Since then, the company has doubled down on live programming, producing a slate of high-profile events including NFL games, WWE, and the Mike Tyson versus Jake Paul fight. Stone said it required not just new technology but a cultural shift within engineering and operations teams.
“Live might look easy. It’s not easy, especially not at the scale that Netflix has around the globe,” she said.
Delivering live events required tighter integration between tech and production teams, new workflows and rapid learning. Stone recalled being in the launch room with engineers in Los Gatos, California, during the Tyson-Paul fight.
In the hours leading up to the event, Netflix experienced a surge in traffic that temporarily caused streaming delays and access issues for some subscribers. According to reports from ESPN and The Athletic, viewers in multiple countries encountered error messages, were unable to log in, or experienced lagging streams. The platform eventually stabilized, but the early hiccups drew scrutiny across sports media and social platforms.
Stone said the moment tested the engineering team’s resilience and preparedness. She joked that while she practised deep breathing in the back of the room, the team calmly executed what she described as one of the most stressful technical challenges she’s ever witnessed.
After that event, she said engineers stayed up all night to reflect on what they had learned, and by the next morning, they had a three-week plan to improve performance ahead of the next live broadcast. Stone said the sense of ownership and accountability in those moments is what sets the team apart.
“You can be sure something is going to go wrong,” she said. “And it’s very hard to predict exactly what that’s going to be.”
Leaders launching new capabilities can’t rely on perfect execution, she stressed. It’s about fostering a culture where teams are empowered to learn fast and iterate under pressure.
AI is a tool, not a replacement
Stone pushed back on the notion that artificial intelligence is fundamentally new to Netflix. She said the company has used machine learning for years, especially in personalization and content discovery.
“Netflix has been using machine learning and AI since well before I arrived at the company,” she said. “It’s really the backbone of how we think about personalization.”
What’s changing is the emergence of generative AI, which enables creators to produce images, video, and text from scratch.
Stone said Netflix sees these tools as creative enablers, not replacements. She cited use cases where showrunners might want to visualize a world or a character in early development, and where generative tools could accelerate or enhance that process.
“The key is that human creativity is still at the centre of the content that gets created,” said Stone. “As these tools are useful for the filmmakers or the show creators that we’re working with, we want to be providing those tools to creators so they can leverage them.”
She also reflected on how to navigate decision-making during uncertainty, drawing on her background in economics and data to describe a strategy that balances curiosity and pragmatism.
“You have to be able to look at data points that look a little bit irregular, and think: is that something that’s changing? Is that something we should be responding to?” she said. “That’s where the judgment comes into play.”
As Netflix moves further into live events, gaming, and other new formats, Stone said risk will remain part of the equation, guided by data, context, and a readiness to respond when things don’t go as planned.
“There should be some things that have very low probability of being successful, but if they are, would be huge for members, for the business, for creators,” she said. “And then there should be some things where we say there’s very high probability this will be successful, and it’s going to add some value. If we have all sure bets, I feel like we’re not taking enough risk.”
