How Knowledge Workers Should Start AI Coding

The Shift to AI-Assisted Building

AI coding has moved beyond the domain of software engineers. Knowledge workers are increasingly using AI to build custom software that solves specific, recurring problems. Data from enterprise usage shows that non-engineering functions—like finance, sales, and legal—are adopting AI coding tools at a significantly higher rate than in previous months. The goal is not to become a software engineer, but to compound professional advantages by building tools that automate or improve existing workflows.

Three Core Build Patterns

To identify what to build, categorize your work into one of three patterns:

1. Automation (Same job, same output): Replacing manual, repetitive tasks like file renaming or template filling. The receiver of the output should not notice a difference if the tool breaks.
2. Upgrade (Same job, new output): Transforming static deliverables into interactive assets, such as turning a PDF report into a live dashboard or a status email into a self-service page.
3. Invention (New job, new output): Creating capabilities that were previously impossible, such as monitoring hundreds of sources or performing large-scale data analysis that would be impractical to do manually.

Delivery Classes and Durability

Software should be built with a specific level of durability in mind. Prototypes are disposable and meant to test ideas. Personal Software is reliable enough for you or a small team but can sacrifice polish or UX. Production-grade software is for external users and requires higher security and stability. Finally, Products are built for the open market. Understanding these tiers prevents over-engineering; you don't need a production-grade system for a task that only you perform.

Identifying Opportunities

Look for software-shaped problems in six areas: presentation (interactive explainers), content (automated pipelines), data (dashboards), documents (batch processing), inbox (intake systems), and admin (lookup tools). Before building, always check if an existing tool or API can solve the problem more efficiently. Building should be a strategic choice, not a default reaction to every inconvenience.

Key Takeaways

• Don't try to become a software engineer; focus on building tools that solve your specific, recurring work problems.
• Start with the Automation pattern for tasks where you already know what "correct" output looks like.
• Use Prototypes to validate if AI can handle a task before investing time in building a durable pipeline.
• Prioritize Upgrade projects to turn routine deliverables into high-value assets like dashboards or interactive tools.
• Always evaluate existing SaaS solutions before building; only build when a custom solution provides unique value that off-the-shelf tools cannot.
• Accept that some projects will be disposable; the cost of building is now low enough that "throwaway" software is a viable strategy.