Sonali Varde on AI and the Engineering Manager Role

The podcast discusses the growing impact of AI on engineering management and software development, focusing on how AI tools are transforming operational workflows, reducing manual tasks, and improving efficiency. AI is being used to automate routine work such as data summarization, dashboard insights, bug triaging, and meeting preparation, allowing managers and engineers to focus on higher-level strategy and decision-making. The distinction between AI "skills" - reusable components for specific tasks - and more complex, autonomous "agents" is explored, with examples like AI-powered bug bash automation demonstrating how teams are streamlining processes.

While AI accelerates prototyping, coding, and cross-team collaboration, it also introduces new challenges. Code reviews have become a bottleneck due to faster development cycles, and there is increased pressure on senior engineers to maintain quality. AI can produce misleading or incorrect outputs, requiring human expertise in debugging, systems thinking, and first principles reasoning to validate results. The discussion also covers shifts in team dynamics, hiring, and onboarding, with an emphasis on adaptability and AI fluency. Metrics like token usage, PR reverts, and OKRs are used to measure AI adoption and productivity, while traditional Agile practices evolve to accommodate more integrated, AI-augmented workflows. Ultimately, AI is seen as a tool that enhances human capabilities but does not replace the need for judgment, ownership, or specialized expertise.

• What if you automated your most time-consuming operational task using AI agents this week?

  • Move: Identify one repetitive workflow (e.g., bug triage, log analysis, dashboard reporting) and build or configure an AI agent (via tools like Cursor, GitHub Copilot, or custom scripts) to automate at least 70% of its execution.
  • Why Now?: AI tooling (e.g., Cloud Code, Gemini, Replit) has matured in the last 10 months, and early adopters are already seeing 3X productivity gains - especially in reducing manual data handling and triage time.
  • Expected Upside: Free up 5 - 10 hours/week for higher-leverage work like prototyping or architecture, and reduce error rates in routine tasks by standardizing AI-driven outputs.

• What if you used AI to create a working prototype instead of a pitch deck for your next idea?

  • Move: Take one product or platform idea you've been conceptualizing and use AI-assisted development (e.g., Replit for mockups, Copilot for CRUD APIs, Gemini for documentation) to ship a minimal interactive prototype in under 5 days.
  • Why Now?: AI has lowered the cost of prototyping - teams now create mini-versions instead of spending weeks on presentations, accelerating alignment with stakeholders and reducing persuasion overhead.
  • Expected Upside: Faster validation of ideas, improved cross-functional buy-in, and earlier user feedback - turning abstract concepts into actionable experiments that drive OKR progress.

• What if you restructured your weekly workflow review around AI-generated insights and token usage metrics?

  • Move: Replace your current status tracking (e.g., Kanban, story points) with a weekly audit of AI activity (token usage, PR reverts, trunk blockages) and use AI to summarize bottlenecks and suggest process improvements.
  • Why Now?: Traditional metrics like velocity are becoming obsolete; token and throughput metrics (e.g., experiments shipped, incidents reduced) reflect real AI-augmented performance and expose new bottlenecks like code review queues.
  • Expected Upside: Shift focus from output tracking to impact measurement, uncover hidden inefficiencies (e.g., review latency), and align personal workflow optimization with team-wide AI adoption trends.

• Implement AI agents to automate repetitive workflow tasks such as bug triaging, JIRA logging, or dashboard updates, starting with one high-frequency task to measure time savings.
• Conduct a personal audit using the Eisenhower matrix to identify low-value, repetitive tasks and develop or adopt AI skills (e.g., automated summarization, data extraction) to offload them.
• Integrate AI tool proficiency into onboarding plans, expecting new solo developers to use core AI tools within 60 days and automate at least one small workflow by 90 days.
• Replace traditional velocity metrics like story points with outcome-based tracking such as shipped features in production and measurable system improvements, aligned with OKRs.
• Regularly validate AI-generated outputs (code, debug suggestions, designs) using first-principles and systems thinking, especially in critical areas like architecture and data policies, to avoid costly rework.