You’re not alone—many teams pivoted fast to agent-written code, and it can feel like craftsmanship no longer matters. A few concrete moves:
- Have a candid 1:1: say you’re misaligned with the process, but propose owning guardrails leadership cares about—reliability, security, test coverage, CI policies, prompt/eval hygiene. Suggest measuring outcomes beyond velocity: defect escape rate, change failure rate, MTTR, SLOs, incident cost.
- Differentiate where agents are weak: ambiguous requirements, system design, debugging gnarly prod issues, performance tuning, threat modeling, compliance. Volunteer for those areas.
- Use AI defensively: generate tests, fuzzers, benchmarks, docs, migrations; prune agent output; write prompts/evals to reduce rework and incidents.
- Protect yourself: keep a brag doc with quantified impact, network for internal transfers, and quietly explore roles that still value rigor (fintech, healthcare, infra, aerospace, devtools).
- Set a 60–90 day window. If nothing changes, execute an exit plan rather than burn out.
It’s okay to be disillusioned. Your edge now is owning quality, risk, and outcomes—things the org can’t ignore even when throughput is cheap.