Which Developer Roles Survive the AI Transition?

The Framework

We've developed a simple framework for evaluating AI risk to any tech role. Ask two questions:

1. Can the work be clearly specified? If you can write a perfect prompt for it, AI can do it. 2. Does the work require context AI can't access? Customer relationships, proprietary business logic, real-world constraints — these are human territory.

The Risk Tiers

CRITICAL RISK (50%+ reduction expected by 2027) - Junior software developer - QA tester (manual) - Technical writer (API docs, code docs) - Tier-1 customer support - Basic DevOps (deployment, monitoring) - Data entry / data cleaning

These roles are primarily execution. The specification is clear, the context is available, and AI is already doing them at scale.

HIGH RISK (25-50% reduction expected) - Mid-level software developer (CRUD apps, standard features) - SRE / on-call incident response - Security analyst (log analysis, alert triage) - Database administrator (routine optimization) - QA automation engineer

These roles have judgment components, but much of the day-to-day work is pattern-matching that AI excels at.

MODERATE RISK (10-25% reduction expected) - Senior software engineer - Solutions architect - Engineering manager - Product manager (technical) - DevRel / developer advocate

These roles require cross-functional coordination, stakeholder management, and contextual judgment that AI can't fully replicate.

LOW RISK (stable or growing) - Staff+ engineer (system design, tech strategy) - Security engineer (adversarial thinking, novel threats) - ML/AI engineer (obviously) - Platform engineer (internal tooling, developer experience) - Domain specialists (healthcare, finance, etc.)

These roles either work on AI itself, require adversarial thinking AI can't do, or depend on context and relationships AI can't access.

The Uncomfortable Middle

The biggest danger zone is the middle of the stack: mid-level developers with 3-7 years of experience who are too expensive to be juniors but haven't developed the system-design skills of seniors.

Companies are discovering they can skip this tier. AI + seniors + juniors (where juniors validate AI output rather than write code) is becoming the new team structure.

What's Actually Happening

Companies aren't announcing "we're replacing developers with AI." They're: - Not backfilling roles when people leave - Combining teams and expecting AI to fill the gap - Restructuring without layoff announcements - Hiring senior roles and eliminating junior ones

The numbers show up in aggregate job postings and layoff data, but the narrative remains "AI augments, not replaces." Don't believe the narrative. Believe the data.

What To Do

If you're in a high-risk role: 1. Move toward the specification side (PM, architect) or the novel-problem side (security, ML) 2. Develop domain expertise that AI can't easily acquire 3. Build relationships and organizational knowledge that make you irreplaceable 4. Learn to work with AI effectively — become the human in the loop

The transition is happening faster than most 2023 predictions suggested. If you're going to move, move now.