Product Manager vs Software Engineer

External PM hiring contracted ~14% in 2025, but the function has not shrunk — companies are concentrating spend on proven mid-to-senior practitioners and paying them more: median PM salaries rose ~5% in 2025, outpacing most peer functions. Entry-level and associate PM roles are the hardest hit; roles at that band face both fewer openings and oversupply from experienced candidates willing to step down. Senior and Group PM compensation is where the real movement is — US Group PM new-offer packages rose 25.6% in 2025. Compensation premiums are concentrating visibly in two places: AI-native product roles (PMs owning LLM-powered features or AI infrastructure products) and PMs with hard quantitative and experimentation skills. Tooling has moved fast: Notion AI, Productboard, and Amplitude are now table stakes for discovery and prioritisation workflows; PMs who cannot interpret A/B test results or write a coherent SQL query are increasingly filtered out at screen stage. The role is not being automated — but the surface area of a single PM is expanding as AI absorbs lower-order execution tasks, raising the bar for strategic and commercial judgment.

The software engineering market has stabilised well below its 2021–2022 peak. Entry-level postings are down 28% from those highs and have not recovered, with AI-driven productivity gains reducing junior headcount at large tech companies — new graduates represent only 7% of Big Tech hires. Mid and senior engineers with specialisation remain in demand: AI/ML engineering, cloud architecture, and security engineering are the three sub-disciplines with active growth; all other tracks face flat or declining posting volumes. AI skills now appear in 42% of software job descriptions, up from 8% in 2022, and engineers who demonstrate AI tool proficiency are securing roles measurably faster. Hiring demand has shifted sectorally — financial services and industrial automation are among the strongest employers, while marketing-tech engineering is contracting sharply. Claude Code has overtaken GitHub Copilot as the most-used AI coding tool as of early 2026, with 75% of engineers now using AI for at least half their work. The premium in compensation is concentrating in engineers who pair strong system design fundamentals with the ability to operate effectively with AI tooling — not in those who rely on it as a crutch.