How Block is becoming the most AI-native enterprise in the world | Dhanji R. Prasanna
How Block is becoming the most AI-native enterprise in the world | Dhanji R. Prasanna
How Block is becoming the most AI-native enterprise in the world | Dhanji R. Prasanna
In this conversation, Dhanji R. Prasanna, CTO of Block, shares insights from leading one of the most AI-native large companies today. From the creation of an internal AI agent that’s reshaping workflows to the organizational shifts that have amplified its impact, he reveals how AI is redefining productivity across engineering and non-technical teams alike.
Dhanji discusses how Block's internal AI agent, Goose, saves employees 8–10 hours weekly by automating routine tasks like report generation and code patching, with even non-engineers building tools rapidly. The company measures AI gains across teams, finding that while current efficiency is already significant, future improvements are expected as tools mature. Surprisingly, organizational structure—shifting from a GM model to a functional, tech-first setup—has had a greater productivity impact than AI itself. Dhanji emphasizes that leadership must use AI tools firsthand to drive adoption and that human judgment remains critical for taste and alignment. He also reflects on past failures like Google Wave, stressing that product success hinges on solving real user problems, not technical elegance. Controlled chaos, narrow scoping, and questioning assumptions are key to innovation. Ultimately, AI amplifies productivity, but lasting impact comes from cultural and structural choices.
07:33
07:33
Writing the AI manifesto directly led to becoming CTO
10:51
10:51
Struggling companies should view themselves as technology-building entities.
12:06
12:06
AI-native teams use Cursor to build software with little manual coding
15:25
15:25
Engineering teams using Goose save 8–10 hours per week
20:18
20:18
Current AI efficiency is the worst it will ever be, pointing to future gains
28:11
28:11
Goose watches engineers' screens and anticipates tasks like opening PRs
35:20
35:20
Goose succeeds on about 60% of well-described features without human intervention
37:42
37:42
Humans are needed to anchor AIs and prevent them from going off-script.
50:57
50:57
Functional org structure is more impactful for productivity than AI adoption
53:45
53:45
Use AI tools like Goose daily to understand how your workflow can change
55:15
55:15
Goose used AppleScript to collate therapy receipts from multiple forms into one Apple Notes entry.
58:01
58:01
Goose is free to download and use across Mac, Windows, and Linux, with monetization limited to LLM token usage.
59:59
59:59
Organizational structure shapes system design through Conway's Law
1:01:56
1:01:56
YouTube succeeded with poor code because it focused on user needs
1:04:56
1:04:56
Engineers should be free to experiment if the foundation is stable
1:08:07
1:08:07
Start small and narrow your scope to build useful products
1:13:36
1:13:36
Failures like Google Wave and Google+ taught valuable lessons that led to Cash App's success
1:21:49
1:21:49
Problems seem trivial in hindsight; don't let fear of change hold you back
