Contents
In brief
The AI-in-engineering debate often swings between extremes: developers are obsolete, or AI is just fancy autocomplete. A webinar with leaders from Platform9, Monday.com, and PlayStation painted a more nuanced picture — teams are restructuring the entire SDLC around AI tools, not merely experimenting with chatbots.
What happened
Among panelists, Cursor and Claude Code lead adoption; GitHub Copilot wins where it already fits the git ecosystem. Platform9 tiers tools: UI engineers use Cursor, backend prefers Windsurf, terminal-heavy users run Claude Code. Monday.com standardized on Cursor locally and Claude Code for automated SDLC phases. PlayStation relies on Copilot for repository integration.
A major shift is toward agentic engineering: developers become orchestration engineers. Code is less of a differentiator; architecture, product understanding, and system design matter more. One team shipped a high-performance Rust product without prior language experience — AI handled syntax while humans managed constraints.
AI has moved beyond the IDE. Platform9 uses agents for Kubernetes root-cause analysis; with deployment-specific context ("context engineering") they resolve 50–60% of performance and configuration issues. RAG tackles technical debt and documentation; Windsurf Deep Wiki generates live repo docs for onboarding.
Why it matters
For enterprises, security is the primary filter: AWS Bedrock, Google Cloud AI, strict human-in-the-loop — AI-generated code never reaches production without review. AI also removes bureaucracy: meeting summaries, auto-filled 300-question security questionnaires for sales.
Challenges remain. Review bottleneck: AI generates faster than humans review — noise in GitHub, bot comments, critical changes get lost. Creative biasing: accepting an AI suggestion too early narrows solution space. Model diversity: some leaders generate with one model and review with another to catch more errors.
In practice
- Restructure SDLC stages around where agents automate vs. where humans decide.
- Match tools to roles (UI / backend / terminal) — there is no single best pick.
- Invest in agent context: deployments, schemas, runbooks — LLMs fail without it.
- Scale review processes for generated code volume; consider cross-model review.
- For enterprise, use managed platforms with data governance, not consumer services.
- Value architectural skills over typing speed.
Takeaway
The future points to autonomous agents across the SDLC, but success depends on asking the right questions and defining environments for agents. AI is not replacing engineers — it is shifting their role. Full webinar notes are in the original on Dev.to.