AI for Engineering Velocity: A Non-Engineer's Guide
The landscape of software development is undergoing a seismic shift, propelled by the accelerating capabilities of artificial intelligence. This transformation is so profound that some industry leaders have issued stark warnings: adapt or be left behind. GitHub’s former CEO, for instance, famously challenged engineers to embrace AI in their daily workflows, suggesting those unwilling might need to seek new careers – a sentiment ironically underscored by his own departure from the company just weeks later. Yet, his observation about developers’ journey through AI adoption, from “dabbling skeptic” to embracing a “transformational” shift, rings true for many.
Indeed, the numbers underscore this paradigm change. A recent survey revealed that a striking 58% of engineers now believe AI can write code more effectively than most humans, a figure poised to climb as AI tools become increasingly sophisticated. However, AI’s impact extends far beyond mere code generation. It is reshaping virtually every facet of the software engineering lifecycle, from crafting comprehensive documentation and generating robust unit tests to streamlining debugging processes and revolutionizing code reviews. Companies like Anthropic, with their Claude Code, are at the forefront of demonstrating these expansive opportunities for AI augmentation.
For product managers and other non-engineers, understanding these shifts is crucial. While it might seem counterintuitive for those outside engineering to delve into the intricacies of “engineering velocity,” these conversations with CTOs are vital for assessing a team’s current AI fluency and identifying untapped opportunities to enhance development processes. A strategic approach to AI integration can yield significant improvements, touching on critical areas such as onboarding new hires, managing technical documentation, optimizing code review cycles, and fostering a culture of AI adoption within engineering teams. There are even compelling examples, such as one company that leveraged AI to eliminate massive technical debt, reportedly saving 280,000 hours of development time. Furthermore, adapting procurement processes to allow engineers to experiment with new AI tools is becoming an imperative for sustained velocity.
Consider the onboarding process for new engineering hires, traditionally a time-consuming endeavor. AI tools have dramatically streamlined this. Platforms like Claude Code, with their immense context windows, can scan entire codebases, including project structures and dependencies, enabling new engineers to simply ask questions about how the system works. They can inquire about logging mechanisms, specific function behaviors, or critical edge cases, gaining insights that previously required extensive manual exploration or peer consultation. Anthropic itself leverages Claude Code internally, even recommending the creation of a CLAUDE.md file within repositories. This file serves as a centralized knowledge base for best practices, coding style guidelines, testing instructions, and repository etiquette, ensuring Claude can provide context-aware guidance to developers. Equipped with such tools, new hires can grasp product fundamentals at an unprecedented pace.
Beyond codebase navigation, AI is fostering new learning methodologies. Recent “Study Modes” introduced by platforms like ChatGPT and Google Gemini offer structured, conversational learning experiences, allowing engineers to deepen their understanding of new technologies or brush up on existing knowledge. While not a substitute for formal training, these features provide an accessible way to explore complex topics, such as Retrieval Augmented Generation (RAG), through interactive dialogue. The broader impact of AI-powered onboarding extends to reducing employee turnover. Data from AI company Glean suggests that AI onboarding assistants can slash first-year employee attrition by 30%, with companies like Airbnb and Uber already deploying their own custom solutions. Even generative AI startups like ElevenLabs are utilizing their own products to automate product tours and provide conversational voice agents for new hires, offering instant insights into complex systems like API platforms.
The transformation also extends to the often-neglected realm of documentation. Writing and maintaining accurate technical documentation has historically been a challenging, resource-intensive task. However, new AI-powered tools are emerging, with some innovative startups already counting industry leaders like Perplexity, Replit, and Cursor among their clientele, poised to fundamentally change how technical documentation is created, updated, and managed.