AI applications often rely more on prompts and tool selection than on traditional code. Product managers can now craft prompts themselves, while engineers lean on coding agents like OpenAI Codex to automate boilerplate.12 This shift means PMs and engineers collaborate more closely on specification and implementation.
Simple LLM apps combine a base model with a prompt template. More advanced systems use retrieval augmented generation (RAG) or agents to populate the prompt with context and call tools.34 As these approaches mature, PMs and domain experts can directly adjust prompts to improve user experience, while engineers focus on integration and evaluation.
Because prompts define behavior, the best prompt engineers will build the best AI products. Tools that measure prompt changes, provide intuitive UIs, and track data flow help teams iterate quickly. Over time, the boundary between planning and building will fade as both roles revolve around understanding user needs and refining prompts rather than writing raw code.