Lenny's Podcast: Product | Career | Growth

How Block is becoming the most AI-native enterprise in the world | Dhanji R. Prasanna

October 26, 2025

Key Takeaways Copied to clipboard!

  • Block's transformation into an AI-native company was catalyzed by a shift from a GM structure to a functional organizational structure, allowing for singular technical focus. 
  • AI tools, exemplified by the internal agent Goose, are currently saving highly AI-forward engineering teams 8 to 10 hours per week, with company-wide savings trending toward 20-25% of manual hours. 
  • Successful product development is counterintuitively decoupled from code quality, as demonstrated by YouTube's success despite its perceived poor architecture, emphasizing that solving user problems is paramount. 
  • Successful product development often stems from starting small with experiments (like Goose and Cash App) rather than immediately committing massive resources to a big idea (like Google Wave). 
  • Core leadership lessons include the necessity of creating 'controlled chaos' by providing engineers freedom to experiment while maintaining a stable foundation, and adhering to the principle of 'start small with everything' to avoid boiling the ocean. 
  • In an era of rapid technological change, professionals should focus on what matters to them (like open source and accessibility) and demand that technology serves their important purpose, rather than chasing every new trend. 

Segments

AI Productivity Gains Measurement
Copied to clipboard!
(00:00:00)
  • Key Takeaway: AI-forward engineering teams report saving 8 to 10 hours weekly, establishing a baseline productivity gain that is expected to increase.
  • Summary: Self-reported data shows engineering teams highly utilizing AI save 8 to 10 hours per week, with company-wide manual hour savings trending toward 20-25%. This metric is considered the baseline, as the utility and value of AI capabilities are changing daily. Companies must adapt to ride this continuously improving wave of capability.
Goose Agent in Action
Copied to clipboard!
(00:00:27)
  • Key Takeaway: An engineer’s Goose agent autonomously built and opened a Pull Request for a feature discussed only in Slack/email.
  • Summary: The internal AI agent, Goose, can monitor engineer workflows, including Slack and email discussions about desired features. It can then proactively attempt to build the feature and open a corresponding Pull Request in Git hours later. This demonstrates an advanced level of autonomous workflow integration.
Non-Technical AI Impact
Copied to clipboard!
(00:00:50)
  • Key Takeaway: Non-technical personnel embracing AI agents show the most surprising and significant impact from the tools.
  • Summary: The most surprising beneficiaries of AI agents are non-technical staff who use the tools to optimize their specific daily tasks. This capability allows teams like enterprise risk management to build self-servicing systems, compressing weeks of work into hours, bypassing traditional internal app development queues.
Future of Engineering Work
Copied to clipboard!
(00:01:12)
  • Key Takeaway: Future engineering work will involve LLMs operating autonomously overnight to build anticipated features, moving beyond current interactive coding sessions.
  • Summary: LLMs should be utilized constantly, working overnight and on weekends when humans are idle, to build software proactively. This shift moves away from the current ping-pong style of vibe coding toward greater autonomy and anticipation of human needs. Engineers will increasingly rewrite entire applications from scratch rather than refactoring.
Code Quality vs. Product Success
Copied to clipboard!
(00:01:29)
  • Key Takeaway: Code quality has virtually no correlation with building a successful product, exemplified by YouTube’s success over Google Video.
  • Summary: Engineers often mistakenly prioritize code quality, but product success is driven by solving the user’s problem effectively. YouTube succeeded despite having notoriously poor architecture (storing video blobs in MySQL), while the technically superior Google Video failed. Focus must remain on user needs, not internal technical elegance.
Block’s AI Transformation Context
Copied to clipboard!
(00:05:37)
  • Key Takeaway: Dhanji R. Prasanna convinced CEO Jack Dorsey to prioritize AI via a central ‘AI manifesto’ document.
  • Summary: The CTO wrote an ‘AI manifesto’ to Jack Dorsey arguing for Block to become an AI-native company, which led to his promotion and company-wide transformation. This shift involved moving from a GM structure to a functional organizational structure to drive unified technical depth.
Functional Org Structure Benefits
Copied to clipboard!
(00:09:21)
  • Key Takeaway: Transitioning from a siloed GM structure to a functional organization enabled Block to drive technical depth and AI adoption effectively.
  • Summary: The previous GM structure treated engineering headcount as a commodity, leading to siloed technical strategies across Square, Cash App, and Afterpay. The functional structure centralizes engineering and design leadership, ensuring a singular technical focus, shared tooling, and consistent engineering standards across the company.
AI Tool Adoption Strategy
Copied to clipboard!
(00:54:04)
  • Key Takeaway: Driving AI adoption requires leadership, including the CEO and executive team, to use the tools daily to understand their strengths and ergonomics firsthand.
  • Summary: Leadership must personally use AI tools like Goose every day to learn how workflows can change before attempting to mandate adoption across teams. This direct experience provides more insight than reading thought pieces and helps identify where tools are most efficacious for application.
Goose Agent Deep Dive
Copied to clipboard!
(00:21:39)
  • Key Takeaway: Goose is an open-source, general-purpose AI agent built on the Model Context Protocol (MCP) that gives LLMs ‘arms and legs’ to act in the digital world.
  • Summary: Goose functions as a desktop application that can interface with enterprise tools (like Salesforce or Snowflake) via MCP wrappers, allowing LLMs to manipulate real-world data and systems. It supports pluggable model providers, including Claude, OpenAI, and open-source models. Its open-source nature allows other companies, including competitors, to implement and extend the platform.
Hiring Mindset in AI Era
Copied to clipboard!
(00:44:09)
  • Key Takeaway: The primary shift in hiring is prioritizing a ’learning first’ mindset and eagerness to embrace AI tools over specific, existing AI practitioner skills.
  • Summary: The functional reorganization, not AI productivity alone, has fundamentally changed headcount planning by moving away from viewing engineers as a commodity. New hires are sought based on their willingness to learn and adapt, rather than mastery of current AI tools. Critical thinking and deep technical understanding remain more important than immediate AI proficiency.
Leadership Lessons Learned
Copied to clipboard!
(01:00:07)
  • Key Takeaway: A CTO must deeply appreciate the power of Conway’s Law and proactively carve out time for holistic judgment, as silence often means problems are going unnoticed.
  • Summary: The structure of organizational relationships (Conway’s Law) dictates outcomes, making structural change critical for achieving new results. Leaders must actively seek perspective because organizational silence during good times can mask underlying issues or missed opportunities. CTOs need dedicated time for holistic assessment, which is often neglected in the day-to-day.
Controlled Chaos and Leadership
Copied to clipboard!
(01:07:03)
  • Key Takeaway: Foundational stability must be secured before allowing engineers the freedom to experiment and iterate.
  • Summary: A foundation that prevents major ruptures or financial loss is necessary before allowing engineers the freedom to experiment. Dhanji R. Prasanna previously held the title of ‘mad scientist’ at Square while working part-time on various projects. This role demonstrated the company’s willingness to afford freedom for experimental work.
Core Leadership Lessons
Copied to clipboard!
(01:08:16)
  • Key Takeaway: Leadership success hinges on narrowing scope by focusing only on the achievable task at hand, echoing Carl Sagan’s principle.
  • Summary: A core leadership lesson is to ‘start small with everything,’ avoiding the attempt to ‘boil the ocean to make a cup of tea.’ This principle means narrowing the scope to the immediate, achievable goal. This tenet has been a core philosophy at Block since its early days as Square.
Small Starts Lead to Success
Copied to clipboard!
(01:09:23)
  • Key Takeaway: Major successful projects like Goose and Cash App originated from small, self-driven experiments or hack week ideas.
  • Summary: Goose began as an engineer building a proof of concept on their own time based on a thesis about agents unlocking LLM value. Cash App also started as a hack week idea that grew into a major product. The first public Bitcoin product launch at Block was similarly the result of a small hackathon team effort.
Counterintuitive Product Lessons
Copied to clipboard!
(01:11:22)
  • Key Takeaway: Starting big immediately, as seen with Google Wave, often leads to failure by lacking adaptation to real-world feedback.
  • Summary: It is counterintuitive to immediately commit large resources to a big idea without testing the market first. Google Wave, which involved 70-80 engineers before significant user adoption, exemplifies a product that started too big. Leaders must constantly question base assumptions about whether a product should be built at all.
Learning from Failures
Copied to clipboard!
(01:13:41)
  • Key Takeaway: A career built on product failures provides humility and critical learning, preventing the repetition of past errors.
  • Summary: Dhanji R. Prasanna’s career includes several failures such as Google Wave, Google Plus, and the social networking startup Secret. These failures instilled humility and taught him to be open to critical viewpoints rather than assuming he has all the answers. Cash App became a major success by applying the cumulative learnings from these earlier setbacks.
Focusing on Purpose Over Trends
Copied to clipboard!
(01:15:59)
  • Key Takeaway: In times of uncertainty, optimize for what is meaningful and fun, as technology should serve human purpose, not dictate it.
  • Summary: People should focus on the things that matter to them, such as open source and improving access for everyone. Technology trends are secondary because technology exists to serve us. If a pursuit is not meaningful and fun, one should likely change it.
Lightning Round Recommendations
Copied to clipboard!
(01:18:08)
  • Key Takeaway: Recommended reading should prioritize fiction, classics, poetry, and philosophy over professional self-help books to expand creative thinking.
  • Summary: Dhanji R. Prasanna recommends The Master and Margarita by Mikhail Bulgakov and Tennyson’s poetry for expanding the mind and finding center during uncertainty. Recent enjoyable media includes the visually stunning pulp sci-fi show Alien Earth and the high-quality spy series Slow Horses. The Steam Deck OLED is a favorite gadget because it champions extensibility and user customizability against industry trends toward lock-down.
Motto and Overcoming Fear
Copied to clipboard!
(01:21:55)
  • Key Takeaway: If professional life lacks energy, one must change their situation, trusting that future hindsight will minimize current fears about change.
  • Summary: The motto is to change something if you are not energized by your professional life each morning. Fear of making a change is often mitigated by remembering that monumental problems in the moment appear trivial in hindsight a year later. It is never too late to do something useful for self-improvement.
Final Call to Action
Copied to clipboard!
(01:24:34)
  • Key Takeaway: Listeners should demand more openness from their companies, specifically by defaulting to open source and ensuring benefits extend beyond immediate customers.
  • Summary: People can be useful by demanding better from their employers and teams, asking if projects can default to open source. This is crucial in the AI era where many are locking down emerging platforms into walled gardens. The internet’s promise of open information sharing should be realized by AI for the benefit of all.