In production250K-member electric cooperative Confidential system, demo on request

Billing System Skills for Claude

Turned knowledge that lived in a few people's heads into five installable AI skills, set up in about 15 minutes per person.

Wrote down how the cooperative's billing system works as a set of Claude skills, so any staff member can ask the AI instead of waiting for the one person who knows.

  • 5 skills: collections, invoicing, areas and books, prompt payors, and notifications
  • About 15 minutes of one-time setup per person
  • One private, versioned source of system knowledge

5 skills, about 15 minutes to set up per person, in use by office staff

Team of one· Solo author & trainer

No team, no agency, no hand-offs. I owned this project end to end:

  • Requirements
  • Architecture
  • AI & prompts
  • Backend & APIs
Stack
Claude CodeAgent SkillsPowerShellPrivate GitHubAI Enablement

Where AI does the work: This is AI enablement, not an AI feature: the work is turning tacit system knowledge into skills an assistant can load, and getting non-technical staff to use it as a working tool.

What was broken, and what fixed it

  1. Problem: Knowledge of how the billing system works lived in a few people's heads.

    Fix: What the experts know was written down once, as skills: how collections, invoicing, areas and books, prompt payors, and notifications actually work in this system.

  2. Problem: Staff had to explain the billing system to the AI from scratch every time.

    Fix: The skills install on each person's machine, so the assistant already knows the system when the question comes.

  3. Problem: Most staff had only ever used AI on a website, never as a working tool.

    Fix: A step-by-step guide written for people who have never opened a terminal, plus an installer that does the setup in one run.

Approach

  • Write the knowledge once and version it, so a correction reaches everyone on the next update.
  • Keep it private: access is by invitation, and the repository stays inside the organization.
  • Make the first 15 minutes easy, because adoption is decided there.

What I built

  • Five skills that teach Claude the cooperative's billing system
  • A one-run installer and a plain-language setup guide
  • An invite-only private repository, improved over time

Architecture

AI

  • Claude Code

Logic & APIs

  • Agent Skills
  • PowerShell
  • Private GitHub
  • AI Enablement
  1. 1

    Expert knowledge is captured as skill files in a private, versioned repository.

  2. 2

    An installer script places the skills on each staff member's machine.

  3. 3
    Claude

    Claude Code loads the relevant skill when a billing question comes up.

Outcome

  • In use by office staff, not only by IT.
  • System knowledge has one shared, versioned source instead of a few people's memory.