We coached the CEO of a homecare staffing agency to build his own AI tools. He now ships reliable internal software to his team himself, starting with an app that matches caregivers to shifts by real driving distance.

The challenge

NursePartners is a Philadelphia-area homecare and nurse-staffing agency that fills open shifts across four counties, often on short notice. It had been mapping caregivers to shifts by broad geographic zones, an approach that was inefficient and imprecise, because zones do not reflect real drive time and schedulers routinely reached the wrong caregivers. Its CEO, a hands-on operator with a data analytics background, was building an internal app himself to rank caregivers by actual driving distance instead. He wanted to move faster and to put real structure around how he uses agentic coding: a way to build without the fear of breaking the live app, and a safety net he could count on as more of the team came to depend on it.

What we built together

• Coached over 30 business days so the skill stays in-house, for this project and the ones that follow, instead of a dependency on us
• Set up a safe workflow: branches and pull requests, automated checks that gate every deploy, and custom domains, so a bad change never reaches the live app
• Showed how to code safely, with API keys and secrets stored securely instead of sitting in the codebase
• Moved development from Claude chat into Claude Code, building directly in the real codebase
• Taught how to match Claude Code's modes, models, and effort levels to the task, along with other optimizations that speed up the work

The results

• The CEO ships changes to a live, in-use tool himself, safely, instead of pushing straight to production and hoping it holds
• The app ranks available caregivers against ~14,000 open shifts by real driving distance, filtered from a 1,566-person staff directory, replacing the zone broadcasts that had been cannibalizing the agency's own bookings
• The scheduling team, the daily users the tool was built for, went from frustrated with the old system to satisfied with the app
• He folded three separate tools (birthday reminders, retention stats, a healthcare tracker) into the app, with four to seven more planned
• The capability is durable: he keeps building the tools his team needs at the speed and cost of a solo AI-assisted developer

Built with