No hiding from AI: 90% of software developers now use AI, warns Google executive

"There's no hiding from AI": What product teams need to do now

Google's senior director of product, Ryan J Salva, is blunt: "There's no more hiding from AI if you're a software engineer." AI is now embedded in the build process - from code generation and reviews to documentation and test creation.

His point lands alongside a Google Cloud finding that AI adoption among software professionals has hit 90% worldwide, up 14% year over year. Developers report spending about two hours per day with AI tools. As Nathen Harvey from Google's DevOps Research and Assessment put it, adoption is so common it's like asking, "Are you using a computer at work?" Source.

What changes on your team

• Less raw coding, more product design, architecture, and problem-solving.
• Specs convert to working prototypes faster, closing the gap between PM planning and engineering execution.
• More contributors can participate in software creation and deployment - not just core engineers.
• Code literacy remains non-negotiable. As Salva warned, "You are going to be entirely unsuccessful if you cannot read the language."

How product leaders should adapt now

• Make AI the default workflow. Treat AI tools as part of the IDE: code generation, test scaffolding, doc drafts, and code review suggestions.
• Rewrite PRDs for machine consumption. Use precise acceptance criteria, example inputs/outputs, and edge cases to feed models reliably.
• Shift team composition. Add platform engineers and developer productivity roles to build templates, guardrails, and evaluators.
• Create an AI review stage. Add checkpoints for licensing, security, and quality on AI-generated code before merging.
• Track the right metrics. Lead time to prototype, test coverage, escaped defects, mean time to recovery, and model-assisted throughput per dev.

Spec-to-prototype in a day: a practical flow

• Write a behavior-first spec with examples and constraints.
• Generate a scaffold (service, endpoints, types) and unit tests with an AI pair-programmer.
• Run tests; fix with AI suggestions; iterate until green.
• Spin up a sandbox preview; collect PM/design feedback the same day.
• Gate with security checks, dependency and license scanning, and performance smoke tests.

Guardrails you need in place

• Ownership and review. Every AI-suggested change must have a human owner and reviewer.
• IP and licensing. Scan generated code for restricted licenses; log provenance of suggestions.
• Data boundaries. Keep proprietary code and secrets out of prompts; use approved connectors and masking.
• Quality and drift. Maintain test suites, linters, and static analysis; monitor regressions introduced by AI changes.

Skills roadmap for engineers and PMs

• Code reading across languages. You don't need to write every language fluently, but you must read and reason about it.
• Prompt patterns for software work. Specs-to-code, tests-from-specs, refactor-with-constraints, and "explain this diff."
• System design with AI in the loop. Clear module boundaries, idempotent services, and testable contracts.
• Evaluation practices. Golden tests, seed repos, and benchmark tasks to compare AI tools and prompt templates.

Org-level moves for Product Development

• Codify AI contribution rules in your SDLC and contribution guidelines.
• Create reusable templates: service scaffolds, test harnesses, PRD schemas, and prompt libraries.
• Stand up an internal "AI enablement" squad to own tooling, training, and vendor governance.
• Pilot with low-risk modules, then scale based on measured gains and defect rates.

Looking to upskill your team's AI fluency for coding and product workflows? Explore focused paths here: AI courses by job.

The takeaway

AI is shifting engineers from typing code to designing systems and solving product problems. The barrier to contribute is lower, but the bar for judgment is higher. Or as Salva put it, "The end product was never just code." Now it's your process, your prompts, and your product decisions - shipped faster, with discipline.