The Support Machine: How to Deploy AI in Customer Support Without Breaking Trust is both a full manuscript and an executable AI-agent skill.
The manuscript is the story and philosophy. The skill is the operating system: it helps a reader plan, audit, launch, and operate an AI support migration.
- Read the book online
- Download the PDF
- Use the skill
- Read the Markdown source
- Agent-readable llms.txt
Use this skill when you need to:
- Diagnose whether AI support is actually the right answer
- Evaluate AI support vendors
- Design the support machine architecture and guardrails
- Run a shadow test before launch
- Build a zero-hour launch runbook
- Review the first 30 days after launch
- Define the post-launch operating model
Core thesis:
AI support is not a chatbot project. It is an operating model migration.
.
├── skills/the-support-machine/
│ ├── SKILL.md
│ ├── references/
│ │ └── book.md
│ └── templates/
│ ├── ai-support-rollout-plan.md
│ ├── first-30-days-review.md
│ ├── launch-day-runbook.md
│ ├── operating-model.md
│ ├── shadow-run-scorecard.md
│ └── vendor-scorecard.md
├── manuscript/
│ └── The Support Machine - manuscript v0.2.docx
├── scripts/
│ ├── build_site.py
│ └── export_docx_to_markdown.py
├── docs/
│ ├── assets/
│ │ └── the-support-machine-social.png
│ ├── index.html
│ ├── book.html
│ ├── skill.html
│ └── the-support-machine-v0.2.pdf
├── LICENSE.md
├── LICENSE-MANUSCRIPT.md
├── LICENSE-SKILL.md
└── evals/
└── evals.json
Copy or symlink the skill folder into your local skills directory:
mkdir -p ~/.codex/skills
cp -R skills/the-support-machine ~/.codex/skills/Then ask:
Use The Support Machine to build our AI support rollout plan.
Or:
Use The Support Machine to audit this AI support vendor proposal.
Zip the skill folder:
cd skills
zip -r the-support-machine.skill the-support-machineUpload the-support-machine.skill in Claude.ai under Skills.
For agents that can read skills from a GitHub repo, point them at:
skills/the-support-machine/SKILL.md
We are choosing between two AI support vendors. Build a scorecard and identify hidden risks.
Help us plan a shadow run before exposing the AI support machine to customers.
Create a launch-day runbook for our AI support rollout.
We launched last month. Build a first-30-days review. Some metrics are missing, so mark what we still need.
This repo follows common patterns from open-source skill repositories:
- A compact
SKILL.mdwith strong trigger metadata. - Long source material kept in
references/. - Reusable outputs kept in
templates/. - Optional scripts for deterministic maintenance.
- Eval prompts in
evals/to test whether the skill behaves correctly.
Reference projects reviewed:
Pull requests are welcome for typo fixes, clearer examples, stronger templates, and better distribution ideas.
For manuscript edits, open a pull request against skills/the-support-machine/references/book.md and explain the intended manuscript change. For skill behavior, edit SKILL.md, templates, or evals directly.
Please preserve the book's metaphor system and do not turn the voice into generic AI-transformation copy.
This repository uses a split license:
- Manuscript and book-format outputs: Creative Commons Attribution 4.0 International
- Skill, templates, scripts, evals, and repo docs: MIT License
See LICENSE.md for the file-by-file split.
Website: barnir.co
Email: assafbar@gmail.com
This is a personal project, not an official Sentry support channel.
Public v0.2.
The manuscript is closed as a document, but the story is not. This repo is the path from static book to living skill.
