Lenny's Podcast: Product | Career | Growth

Inside Google's AI turnaround: The rise of AI Mode, strategy behind AI Overviews, and their vision for AI-powered search | Robby Stein (VP of Product, Google Search)

October 10, 2025

Key Takeaways Copied to clipboard!

  • Google's recent AI success, evidenced by Gemini hitting the top app spot, stems from a compounding effect of focused effort, urgency, and close collaboration between product and research teams, rather than a single organizational change. 
  • The core Google search experience remains vital for a vast array of specific needs, while AI is proving to be an 'expansionary' force, fulfilling new curiosity and driving growth through multimodal experiences like Google Lens. 
  • Embodying 'relentless improvement' requires being the physical manifestation of constant effort toward positive productivity, coupled with a perpetual dissatisfaction with the current state of the product, pushing for continuous betterment. 
  • Successful product building relies on four core principles: deeply understanding people's 'jobs to be done' (including emotional needs), applying analytical rigor to root causes, designing for clarity over cleverness, and maintaining humility. 
  • The long and iterative development of Instagram's Close Friends feature, which initially failed due to confusion over its purpose and design clarity, highlights the necessity of sustained investment and deep user understanding for transformative products. 
  • Intense curiosity, the desire to constantly question and chase down 'why,' is a critical trait for success in product development and understanding the world, and AI can serve as an ultimate engine to fuel this curiosity. 

Segments

Google’s AI Momentum and Urgency
Copied to clipboard!
(00:00:00)
  • Key Takeaway: Google is experiencing an internal shift characterized by intense focus and urgency to deliver consumer AI products quickly, leading to recent successes like Gemini topping app charts.
  • Summary: The perception of Google stagnation has shifted due to a new sense of focus and urgency internally. This momentum is attributed to the compounding effect of relentless monthly improvements in models and products. The tipping point has been reached where AI models can now truly deliver powerful consumer experiences.
AI Expansion vs. Search Replacement
Copied to clipboard!
(00:06:08)
  • Key Takeaway: AI is expansionary for Google Search, fulfilling an ever-wider set of user curiosities rather than replacing the core search function for established needs.
  • Summary: Users rely on Google Search for a vast range of specific tasks like getting phone numbers or directions, which foundational needs are not changing due to AI. AI is creating growth by enabling more complex questions and visual searches via tools like Google Lens. Visual searches through Lens are growing at 70% year-over-year on an already massive scale.
AI Mode Product Structure
Copied to clipboard!
(00:09:41)
  • Key Takeaway: AI Mode integrates three core components of next-generation search: AI Overviews for quick answers, multimodal visual search via Lens, and a dedicated frontier experience for deep, conversational querying.
  • Summary: AI Mode is designed as an end-to-end frontier search experience leveraging state-of-the-art models for back-and-forth conversation. It taps into Google’s rich data graphs, including 50 billion products in the shopping graph and 250 million places in Maps. This experience is increasingly integrated, allowing follow-up questions from AI Overviews or Lens to seamlessly transition into the deeper AI Mode.
The Evolution of Search Interaction
Copied to clipboard!
(00:13:32)
  • Key Takeaway: The evolution of search is cycling back to natural language questioning, mirroring early concepts like Ask Jeeves, as users now input long, complex queries directly into Google.
  • Summary: Users are increasingly using natural language to ask long, complex questions, such as detailed date night planning queries, which Google’s AI is now equipped to handle. This shift contrasts with the previous era where users optimized queries for keywords to satisfy the search engine. The ability of AI to handle complexity validates the need for this more conversational interface.
SEO in the Age of AI Overviews
Copied to clipboard!
(00:15:30)
  • Key Takeaway: Content creators aiming to appear in AI Overviews should focus on creating extremely helpful, original content that satisfies user intent and cites sources, as the underlying AI research process still relies on core search quality signals.
  • Summary: AI models use ‘query fan out,’ executing dozens of background searches to gather content before constructing a response. The quality of the resulting AI answer is heavily influenced by the quality of the source content, judged by established guidelines like satisfying user intent and providing sources. Creators should focus on content that addresses the complex, advice-oriented questions people are now asking AI.
Building Successful AI Products
Copied to clipboard!
(00:18:50)
  • Key Takeaway: A key recent development in building successful AI products is the increasing ability to steer and utilize sophisticated models through natural language instructions alone, reducing the need for heavy prompt engineering or post-training.
  • Summary: The interface for steering AI is becoming highly human-like, allowing users to issue complex instructions without needing specific ‘incantations’ or hacks. Models are increasingly encoded to use reasoning, tool use, or code execution based on natural language descriptions of APIs or data schemas. This democratization means less heavy-duty fine-tuning is required to achieve sophisticated outcomes.
Relentless Improvement Philosophy
Copied to clipboard!
(00:21:31)
  • Key Takeaway: Product success is driven by embodying ‘relentless improvement,’ which means being the harshest critic of your own work, motivated by a deep desire to make the world better, not just achieving metrics.
  • Summary: This philosophy requires two components: relentlessness (exerting complete effort toward positive productivity) and a commitment to always making things better, never being content. This mindset aligns with the idea of questioning accepted norms (like Tony Fadell’s sticker example) to identify and solve frustrating user problems. The compounding effort from this dissatisfaction eventually leads to a tipping point where the product becomes net useful.
Lessons from Instagram Stories Launch
Copied to clipboard!
(00:30:10)
  • Key Takeaway: When introducing a new format like Stories into an established product like Instagram, success depends on making it complementary by incorporating platform-specific primitives and addressing user frustrations that the original format ignored.
  • Summary: It is crucial to understand the core essence of the product and learn from successful formats elsewhere, recognizing that not every great idea will be invented internally. Instagram made Stories its own by adding unique creative tools and addressing pain points, such as allowing uploads from the camera roll and enabling users to pause stories. New formats must feel coherent yet distinct, respecting user expectations while solving use cases the core product cannot address.
Growth in Established Products
Copied to clipboard!
(00:35:28)
  • Key Takeaway: Driving significant growth in mature products requires humility, observing where user needs are shifting (e.g., from public feed to private sharing), and applying first-principles thinking based on the user’s ‘job to be done’ to identify new, complementary formats.
  • Summary: Product leaders must remain humble servants to the user base, constantly observing shifts in user needs, as market conditions change rapidly. By focusing on the underlying job a user is hiring the product to do, teams can avoid incrementalism and instead build new, complementary features that expand the product’s utility. Successful expansion often involves clearly defining the new feature’s distinct attributes while ensuring it remains coherent within the existing system.
Balancing Optimization and Innovation
Copied to clipboard!
(00:40:08)
  • Key Takeaway: Resource allocation between optimizing existing products and betting on new innovations should be guided by tracking S-curves and expected value, shifting investment away from features approaching diminishing marginal returns.
  • Summary: Mature systems show diminishing marginal returns where adding more resources yields little needle movement, signaling a need to pivot toward new growth drivers. New, successful features create new growth engines where optimization yields significant percentage wins, which is visible in the data. Instrumentation and metrics are essential guides to quantify impact and know when to stop optimizing one area and invest in a new, first-principled bet.
AI Mode Development Timeline
Copied to clipboard!
(00:43:39)
  • Key Takeaway: AI Mode evolved from AI Overviews over approximately one year, driven by observing user behavior (like appending ‘AI’ to queries) and a small team’s conviction that a dedicated, powerful conversational interface was necessary.
  • Summary: The initial driver was the need to answer harder questions that AI Overviews couldn’t handle, leading a small team to prototype a dedicated, blank-screen experience. This prototype quickly demonstrated ‘magic’ by integrating search, reasoning, and multi-turn context, convincing leadership to invest further. The rollout progressed from internal testing to a trusted external tester group before launching publicly via Labs.
Core Product Principles Framework
Copied to clipboard!
(00:52:02)
  • Key Takeaway: Effective product development hinges on deeply understanding user motivation, rigorous problem analysis, and prioritizing design clarity over cleverness.
  • Summary: The framework for successful products involves deeply understanding why people ‘hire’ a product, often requiring interrogation techniques to find the ‘big hire’ moment. Teams must use analytical rigor to dissect metrics drops and find true root causes, exemplified by the initial failure of Instagram’s Close Friends due to mistranslation causing low engagement loops. Clarity in design, like naming a feature ‘AI Mode,’ leverages existing user mental models rather than reinventing standards unnecessarily.
Close Friends Iteration Deep Dive
Copied to clipboard!
(00:57:07)
  • Key Takeaway: The successful iteration of Instagram Close Friends required isolating the core emotional job—achieving connection via DMs from a small group—and simplifying the system around that loop.
  • Summary: The initial version of Close Friends failed because it was confusingly integrated across multiple surfaces (feed, profile) and the user list size was too small (often one or two people) to close the connection loop. The job to be done was emotional connection, not just utility, which was achieved when users added 20-30 people, leading to successful DM replies. Clarity was restored by moving the indicator (green ring) to the outside of the story tray, making the feature discoverable.
Resource Allocation for Breakthroughs
Copied to clipboard!
(01:03:01)
  • Key Takeaway: Building truly transformative products, especially those based on hard technical breakthroughs, often requires significant resources and teams, contrary to the ’lean and scrappy’ startup cult.
  • Summary: While small teams can build impact, teams building products based on major technological shifts often underinvest and give up too early due to insufficient momentum. The decision to scale resources should be based on hitting a conviction moment validated by external usage, not just internal belief. Holding onto small teams too long can lead to slow iteration cycles that would kill a startup, suggesting investment should match the complexity of the required solution.
AI’s Future in Multimodal Inspiration
Copied to clipboard!
(01:06:39)
  • Key Takeaway: The next major evolution of AI utility will be moving beyond text-based tasks into multimodal, inspirational needs, such as visual design and shopping assistance.
  • Summary: AI is moving beyond text modality to assist in visual and inspirational tasks, evidenced by the development of a visual version of AI Mode announced at I/O. This will allow users to generate and iterate on image boards for design inspiration (e.g., office decor) using natural language, potentially disrupting platforms like Pinterest. This shift emphasizes AI’s role in fulfilling inspirational needs alongside core utilities like coding.
Curiosity and Learning in the AI Era
Copied to clipboard!
(01:10:57)
  • Key Takeaway: Intense, persistent curiosity about ‘why’ things are the way they are is the most valuable mindset, which can be amplified by using AI to discover and engage with original, deep-source knowledge.
  • Summary: Embodying curiosity means constantly wanting to know why things are the way they are, which is crucial for understanding users and problems. AI acts as an ultimate curiosity engine, but true learning requires blending AI discovery with engaging in analog, old-school learning like reading original papers or books rather than just summaries. For children, conversational voice AI (like Google Live Search) provides an accessible, natural way to become AI-native learners.
Lightning Round Insights
Copied to clipboard!
(01:15:04)
  • Key Takeaway: Urgency and scrappiness in seizing opportunities, exemplified by securing Justin Bieber’s involvement in the Stamped app, often trump lengthy planning.
  • Summary: Recommended books include ‘Competing Against Luck’ and ‘Design of Everyday Things,’ alongside the fiction ‘Aurora.’ Robby Stein highly recommends the Purple Pillow for its honeycomb polymer technology. The motto ‘Be Curious’ guides his approach to understanding the world and people, and securing Justin Bieber for Stamped involved immediate, urgent action after receiving a one-line reply from his manager.