In a recent LinkedIn post, Melissa Perri discusses a crucial challenge in leveraging Artificial Intelligence for product development: the gap between strategic intent and AI comprehension. Perri, a renowned product strategist, outlines how her team successfully navigated this by transforming abstract strategies into concrete, machine-readable instructions, thereby avoiding common pitfalls of AI implementation.
Perri begins by emphasizing the fundamental requirement for AI to understand explicit instructions, stating:
“Strategy is useless to AI until you turn it into instructions it can actually read.”
She elaborates on the common cycle of AI guessing and human correction, a loop that can be inefficient and frustrating. The core of her solution, as detailed in her post, lies in establishing clear, documented guidelines akin to onboarding human specialists.
From Abstract Strategy to Concrete Instructions
Melissa Perri highlights that simply having a strategy, whether in one’s head or in an inaccessible document, leaves AI operating in the dark. To overcome this, her team moved away from writing a single, monolithic prompt. Instead, they developed a comprehensive set of guidelines, documented and committed to GitHub, ensuring Claude, their AI coding assistant, could access them automatically throughout the website rebuilding process.
Defining Guidelines for SEO and Content
According to Perri, the initial step involved articulating how they discuss their product, including its positioning, unique value proposition, and key messaging. This was translated into specific rules that Claude could apply to each page, supported by a testing framework. “We concentrated on technical SEO a lot here and made sure we were specific, and everything was testable,” Perri notes, underscoring the importance of precise, verifiable instructions.
Ensuring Design Consistency with AI
Perri explains how a detailed style guide, encompassing colors, typography, buttons, and components, was instrumental in maintaining design consistency. Claude was mandated to adhere to these established parameters, preventing the invention of new class names or hardcoding of colors. She recounts a specific instance where a minor but recurring issue was resolved by formalizing it into a rule:
“I kept deleting little uppercase labels above our headings, so it became a rule, and Claude never added one again.”
This systematic approach ensured that corrections were not only made but also learned from, preventing repeated errors.
Establishing Development Protocols
The development guidelines covered the entire lifecycle of feature building, error handling, and the definition of ‘done.’ Perri emphasizes that the AI was programmed to write its own tests for APIs and other critical components, with these tests being re-run before every deployment to prevent regressions. This proactive testing mechanism ensures that previously functional elements do not silently break.
The Power of Living Documentation
A key takeaway from Perri’s experience is the benefit of having these operational rules embedded directly within the codebase. “Because it all lives in the codebase, any agent working in there picks up the rules, and the work that came before, automatically,” she writes. This ensures that context is perpetually available to any AI agent interacting with the system, eliminating the need for manual context-pasting.
Perri reveals the effectiveness of their approach, noting that Claude rated their setup among the top 1-2% of users for the thoroughness of its documented rules, plans, and memory. She draws a parallel between these AI operational strategies and best practices in Product Operations for human teams, advocating for clear documentation, accessibility, and continuous improvement.
“Give your AI the same guardrails you would give a new hire, and put them where the work actually happens.”
Perri concludes by promising a future post that delves into running AI coding agents more like a structured engineering team rather than a simple chat interface.
📝 About This Content
This article is based on insights shared by Melissa Perri on LinkedIn.
📅 Originally posted on July 21, 2026 | View original post on LinkedIn →