The Pragmatic Engineer Podcast: How AI is Changing Software Engineering at Shopify

Guest: Farhan Thawar, Head of Engineering, Shopify
Host: Gergely Orosz
Date: July 2, 2025
Location: LDX3, London

Overview

In this engaging live episode, Gergely Orosz sits down with Farhan Thawar to uncover how Shopify has reshaped its tech culture and operations by going “all in” on AI. The conversation tackles practical implementation, leadership philosophy, AI tool adoption (for engineers and non-engineers alike), organizational changes, and the real cultural impacts of “AI-first” thinking at scale.
This episode is invaluable for software engineers, tech leads, and engineering leaders who want an inside look at how a major tech company manages the transformative impact of AI, both technically and culturally.

Key Discussion Points & Insights

1. Farhan’s Role & Leadership Approach

• Shopify’s culture promotes hands-on, curiosity-driven leadership.
• Farhan is known for getting directly involved— even with tasks outside traditional leadership “swim lanes.”
• Memorable Moment:
  • Farhan personally fixed the Wi-Fi for Shopify’s annual summit, installing 300 access points to improve connectivity, earning the internal title "Chief Wi-Fi Officer."
    • “We don't try to put people into these roles where you are like you're in product, so only think about product or you're in engineering only think about like how the code is written or architecture. We're very much just curious problem solvers.” (02:25)

2. AI Lab Partnerships & Pairing with Researchers

• Shopify actively seeks hands-on collaboration with AI labs (Anthropic, OpenAI, others).
• Farhan describes how he paired for an hour with Anthropic’s applied AI engineers to mutually explore and learn about their new “Claude code” functionality.
  • “I reached out to Anthropic and said, hey, I'd like to pair with one of your applied AI engineers… We can together figure out where we think this product could go.” (07:00)
• Recording and sharing sessions inside Shopify are used for learning and dissemination.

3. Code Red: Tech Debt Reduction

• Shopify ran a 7-month “Code Red” refactoring period to squash tech debt signals (e.g., exceptions, segfaults).
• Used strict metrics (unique exception count, segfaults = 0) to measure progress (08:46–10:45).
• Quote:
  • “We took something between like 30 to 50% of engineering… we just said, these things have to not grow anymore.” (08:46)

4. Shopify’s Journey with AI Coding Tools

• Shopify was the first company outside of GitHub to use Copilot (2021, pre-ChatGPT); gained free early access in exchange for feedback.
• Now, main AI tools for engineers: GitHub Copilot, Cursor (expanding from VS Code focus), and experimenting with cloud code and others.
  • Non-engineers (sales, support, finance) have begun adopting Cursor for custom workflow automation and data access.
  • “The most interesting thing about Cursor is that the growth... is happening a lot outside of engineering and R&D; finance, sales, support.” (12:37)

5. The Rise of “Vibe Coding” & Democratization of Automation

• Non-technical staff are “vibe coding”—building personal workflow software, often via MCP servers and agents, sometimes without engineer assistance.
  • This trend is paralleled to the rise of WYSIWYG tools but with code (14:10).
• “Now it's end of one, right? They're building software for each individual person, not like infrastructure for all people.” (13:57)
• Risks: Large AI-generated PRs create review burden; expectation is that all code submitters truly understand their changes (15:38).

6. SaaS Disruption & Expanding the Software Pie

• AI may enable “personal SaaS,” but Farhan is not worried about traditional SaaS models yet, citing Jevons Paradox—the more software creation is enabled, the more demand grows.
  • “The more we get, the more we want... probably 10,000 times as much [software] as there is now.” (18:04, 19:08)

7. Organizational Changes: AI is the Default

• Internal memo from CEO declared company-wide expectation to use AI tools; performance evaluated as if employees have AI available (20:11).
  • “You don’t have to use AI... but we’re going to expect that your impact is evaluated as if you had the tool.” (20:11)
• Non-R&D areas rapidly picked up LLMs post-memo.

8. Internal Tooling—LLM Proxy & MCP Adoption

• Shopify built an internal LLM proxy to securely manage and expose enterprise AI APIs (LibreChat), handle access tokens, and track usage/costs. Active encouragement of high AI usage, even leaderboard tracking (22:39).
• Wide adoption of MCP (Machine Connection Protocol) for integrating internal/company data into AI tools seamlessly (23:32).
  • “When I talk to people about how to get access to some piece of data..., we quickly will spin up an MCP endpoint for them so that they can just use it.” (21:50)

9. Cost: Open Investment in AI Tools

• Unlike many CTOs, Shopify has no spending limit for engineers/teams on AI tools; they view cost as a function of potential gain.
• $1000/month on AI tools for a 10% productivity gain is considered an “insanely good deal.” (27:19, 28:47)
  • “People are looking at this the wrong way. $1,000 a month is too cheap... you should not be penny-pinching on AI tools because the productivity gains are there.” (27:19)
• Leadership publicly celebrates high AI token use.

10. Intern Programs & Culture Change

• Shopify hiring 1,000 interns per year—10% of total engineering headcount, aimed at infusing “AI-native” talent and accelerating cultural adaptation.
  • “We do it to learn from the interns. That is always the reason.” (34:54)
• Interns and early-career hires are seen as “AI centaurs”—people blending instinctive LLM skills with domain expertise (33:56).

11. Engineering Workflow, Tooling, and Automation

• Internal PM tool “GSD” (“Get Shit Done”), built for Shopify’s unique process.
• AI now drafts weekly project updates from PRs and Slack, but with expectation that leads review/edit for reflection (38:02–38:46).
• Managers/directors use dashboards for team focus, AI adoption, project participation, etc. (39:53).
• Engineers going deeper into “toil reduction,” leveraging AI for tasks like refactoring and debt reduction (43:46).

12. Interviewing Engineering Leaders: Coding Required!

• All engineering directors and VPs must now do a coding interview—with AI tools permitted and even expected.
  • “It is shocking for VPs especially to be like, whoa, there's a coding interview? ...We believe people should be as close to the details as possible.” (40:45–41:50)
• Emphasis is on judgement: assessing if candidates can critique, edit, and reason about AI-generated code as well as prompt it.
  • “If they don't have a copilot, they will lose. But when they do have a copilot, I love seeing the generated code because I want to ask them, what do you think? Is this good code? Is this not good code?” (42:25)

13. Practical Advice and Role Modeling for AI Adoption

• Farhan’s #1 tip: Role model AI adoption—leaders must be hands-on and share their learning and prompts. Internal prompt libraries and sharing use cases are key (44:59).
  • “Role modeling is the best example. I haven't seen anything work better... You have to do it.” (44:59)

Notable Quotes

• On leadership and culture
  • "We're not a swim lane company. We are just curious problem solvers." (02:25)
• On AI tool cost
  • “If I could give you a tool that could make your engineering team more productive by even 10%, would you pay for it? … $1,000 a month is too cheap.” (27:19)
• On making engineers use AI
  • “The expectation is: your impact is going to be evaluated as if you had the tool.” (20:11)
• On interns and cultural change
  • “We do it to learn from the interns. That is always the reason.” (34:54)
• On interviewing leaders with coding + AI
  • “It is shocking for VPs especially to be like, whoa, there's a coding interview?... We want our engineering leaders to be coding as well.” (40:45)
• On who benefits from AI
  • "The camera phone benefited the experts more… AI agents are going to help the best engineers more than the mediocre engineer." (19:08)
• On advice for transformation
  • "Role modeling is the best example. I haven't seen anything work better than role modeling." (44:59)

Timestamps for Key Segments

• Shopify culture & leadership philosophy: 02:25–03:51
• Pairing with AI labs (Anthropic): 06:34–08:10
• Shopify’s “Code Red” and tech debt war: 08:27–10:45
• AI tool adoption journey: 11:05–12:54
• Non-engineers using AI tools (Cursor & MCP): 12:54–14:37
• Vibe coding & democratization: 14:10–15:38
• Should SaaS be worried? Jevons Paradox: 16:28–19:08
• Company-wide AI expectation (internal memo): 20:11–21:06
• LLM proxy & MCP internal infrastructure: 21:40–25:30
• Cost, usage, and productivity philosophy: 27:19–29:33
• Leadership by example; CEO/CTO AI habits: 30:00–31:47
• Intern hiring strategy: 33:16–35:36
• Engineering/project management with AI: 36:37–38:46
• Director/VP hiring & coding interviews (with AI): 40:45–43:24
• Advice for starting AI-first transformation: 44:59–46:07

Memorable Moments

• Farhan personally troubleshooting Wi-Fi at company summit, refusing to silo himself from “low-leverage” jobs (02:25–03:28)
• Democratic approach to tool adoption—if a tool, even expensive, brings productivity, buy it for everyone (27:19–29:33)
• Candidates expected to use AI during director/VP coding interviews; those who don’t end up at a clear disadvantage (42:25–42:29)
• Interns as "AI centaurs," recruited to inject next-gen skills/culture and “change us from the inside” (33:56)

Tone and Spirit

The conversation is open, practical, slightly irreverent, and relentlessly focused on results—not buzzwords. Farhan embodies a hands-on, experimental leadership style that permeates the organization, combining deep technical curiosity with organizational pragmatism.

Fans of The Pragmatic Engineer and anyone interested in pragmatic, concrete approaches to AI integration in software engineering will find this episode valuable and actionable.