The Agentic Era Of Mobile Engineering

Key learnings on how LLMs are transforming engineering teams, the challenges of working with non-deterministic AI, and practical approaches for integrating AI effectively into team workflows.

Transcript

1. The Agentic Era Of Mobile Engineering April 2026 Erik Jhordan Rey, he/him/his Senior Manager, SDE
2. 2 1 Meeting Topic AI is changing how engineering works 2 How We apply AI in our Mobile Team 3 Safe and Smart AI Adoption (Guardrails)
3. AI is changing how engineering works Engineering AI
4. “AI doesn’t just change how we code, it changes how we build software” Engineering AI
5. AI in Requirements & Discovery Accelerating the "pre-code" phase Engineering AI
6. 6 AI speeds up understanding, humans ensure relevance • it helps teams understand problems faster • generate structured requirements • identify risks early • AI supports decisions, it doesn’t replace them AI Capabilities • Set priorities and business context • Domain knowledge is critical • Validate outputs and assumptions • Make final decisions Human Role Engineering AI
7. “human judgment is still essential to make the right decisions” Engineering AI
8. AI in Design & Architecture Enhancing structural decision-making Engineering AI
9. 9 AI amplifies architectural vision and expands design possibilities • Suggest High-level architecture choices • Recommends how to organize modules, classes, or services for maintainability • Generate diagrams, data models, and API contracts • Compare performance, cost, and complexity across options • Highlights potential edge cases, bottlenecks, or architectural risks early AI Capabilities • Choose the right architecture for context • Validate scope, assumptions and constraints • Balance technical and business trade-offs • make decisions based on business needs, team capabilities, and domain knowledge Human Role Engineering AI
10. “Choosing the right architecture still depends on experience, context, and business needs” Engineering AI
11. AI in Coding & Implementation Boosting individual developer output Engineering AI
12. 12 AI accelerates coding, but humans ensure the code is safe, maintainable, and reliable • Automates repetitive code setup, templates, and project structure • Autonomous Code Generation • Assist in Real-Time Coding • Highlights potential errors, suggests fixes, and improves code • Creates test cases to ensure coverage and reliability. AI Capabilities • Review and validate AI-generated code • Debug and fix mistakes • Make architectural and logical decisions • Ensure quality, security, and maintainability • AI writes code, humans ensure it works Human Role Engineering AI
13. “AI tools change the work, but what makes a great engineer remains largely unchanged.” Engineering AI
14. How We apply AI in our Mobile Team Apply AI
15. In our mobile products consumer-facing Apply AI
16. 16 A new experience that helps buyers and renters discover homes through conversation, receive guidance and take real steps such as scheduling tours or connecting with a local agent. “We’re connecting the entire housing journey with AI in a way that hasn’t been possible before,” said Jeremy Wacksman, CEO of Zillow. https://www.zillow.com/news/zillow-debuts-ai-mode/ Zillow AI mode, bringing guided intelligence to every step of the housing journey Apply AI
17. In our engineering workflow How mobile teams build Apply AI
18. 18 • Prompt engineering & AI model interaction • AI-assisted code review & refactoring • AI-driven testing & automation • LLMs • MCPs • Skills • Ethical AI & responsible implementation AI Development AI-powered development tools for Android & iOS AI Development Tools Claude Opus 4.6 Claude Sonnet 4.6 Claude Haiku 4.5 GPT 5.4 GPT 5.3 Codex Composer 1.5 Model Cursor • IDE • CLI Claude Code • TUI • Desktop App Codex • CLI • Desktop App Apple Intelligence • Xcode Environment
19. Piloting Emerging AI Tools for Development Learning what works in practice Apply AI
20. 20 AI Testing Tools AI-Driven Testing • AI Testing tools allow users to write UI tests without any code. • Automate functional testing and edge case detection • An AI agent follows your test steps and interacts with the app as a human would, performing actions and verifying results. • If the AI tester cannot complete an execution step or fails to confirm a verification step, it fails the test and explains why. Pros Cons • Tests written in natural language • Ability to run tests that mirror the user’s actual experience • Tests are easier to maintain and adjust • Visual reports of test results • Tests detail in plain english what went wrong when they fail • Non-determinism causing inconsistent result • Buggy behavior internal to the frameworks • LLM has to reason out every step, costing tokens and time • These tools are still changing and evolving and may be unstable Apply AI
21. Safe and Smart AI Adoption (Guardrails) AI Adoption
22. Set AI Guidelines, Ensure Safety, and Measure Adoption != Impact. AI Adoption
23. 23 • Duration from task initiation to completion • Measures if AI is helping the team deliver faster Cycle Time Change Failure Rate Measuring AI Success in Engineering Acceptance Rate • Check if faster commits create bottlenecks in Code Review or QA • Ensures overall workflow efficiency is improved • Ensures speed doesn’t compromise quality • Tracks errors or bugs introduced by AI-generated code • Frequency of AI-generated suggestions actually used by developers • Indicates practical usefulness, not just availability. Downstream Effects AI Adoption Decide what success looks like for your team!
24. Common Pitfalls to Avoid AI amplifies what’s already there, good or bad. AI Adoption
25. 25 AI is powerful, but safe, responsible use keeps your team efficient and secure • Define ethical standards and AI usage guidelines • AI amplifies workflows, doesn’t fix broken processes • Train engineers to prompt, validate, and guide AI • Avoid unvetted tools that could leak data • Always review AI-generated code manually • Creates test cases to ensure coverage and reliability. AI Adoption • Monitor AI recommendations to ensure alignment with project goals • Document lessons learned to refine AI adoption strategy • Keep an eye on costs make sure the tools are worth it “Don’t just copy-paste AI outputs always verify, because LLMs aren’t always correct or the source of truth”
26. 26 Further Reading • Zillow Mexico Careers • The next chapter of AI in real estate • How Zillow’s new AI mode works • Exploring Generative AI • To vibe or not to vibe • Agentic AI Foundation • Zillow AI mode Engineering AI
27. Thank You!