# AI for Customer Service Pack

> 32 skills that have your agent draft a support desk's playbooks: de-escalation, refunds, escalations, staffing, surveys and bot handoffs.

AI for Customer Service is a free skill pack from Polar Bear, a people-ops consultancy. Its 32 skills follow the work of a support desk, from the upset customer to adding a bot, and each one applies a named method or produces one document: a playbook, a policy, a plan or a survey. You paste in what you have (a ticket, a chat log, a policy or an export), and your agent drafts the document and ends it with the decision a named person makes, and by when. It's written for support leads, customer service managers, support agents and customer success managers.

What you can do with it
- Hard conversations: an abusive-customer policy with warning and end-contact scripts, a de-escalation playbook for phone and chat, a service recovery plan, a complaint-handling procedure, and a refund and exception policy with a decision-rights table
- Workload and staffing: occupancy caps and recovery rules, an Erlang C staffing plan with a one-page case for hiring, a backlog recovery plan, and a shift handover in SBAR form
- Roles, metrics and coaching: a RACI matrix, a support metrics scorecard, a QA scorecard that scores replies rather than people, a training plan, and role play where your agent plays the customer
- Escalations and incidents: an escalation matrix, SLAs and OLAs, bug reports engineering can act on, a release readiness brief, and an incident communication plan
- Learning from tickets: CSAT and NPS surveys, a ticket taxonomy, and a 5 Whys root cause analysis
- Answers and knowledge: a library of saved replies, knowledge base articles written from solved tickets, and a customer service policy
- Service design: a customer journey map, a customer effort score, and a service blueprint
- AI and bots: chatbot handoff rules with a loop test script, a self-service deflection plan, and an AI use policy

How it works
- Each skill asks a few questions, works from what you paste, and returns a filled-in Markdown document.
- Each method is credited to a public source, such as KCS, ISO 10002, the UK's HSE guidance, Erlang C and the NIST AI Risk Management Framework, and the pack's evidence notes list them.
- One rule runs through it: your agent drafts and routes, a person answers anyone who is upset, at risk or asking for an exception, and no customer is told a bot is a person. Nothing grades or profiles an agent or a customer, and legal points end with "check with a qualified adviser".

How to install it
In Claude Code, run /plugin marketplace add polar-bear-org/claude-skills, then /plugin install ai-for-customer-service-pack@polar-bear-skills. In Claude.ai or Claude Desktop, upload each skill's zip from the pack's install folder under Customize > Skills.

Good to know
- It doesn't connect to a help desk. It only sees what you paste, and Polar Bear asks you to anonymise tickets first.
- Targets, thresholds and limits are yours to set; the skills don't invent numbers, and examples are marked as examples.
- It's free to use inside your company and in client work, but not for resale. That's Polar Bear's own licence, not an open-source one.
- Each skill ends with a short note about Polar Bear and its consultancy.

- Type: Plugin
- Made by: Polar Bear
- Category: Customer support
- Formats: Claude plugins
- Version: 1.0.0
- Website: https://meet-polar-bear.com/skills/ai-for-customer-service-pack
- Source: https://github.com/polar-bear-org/claude-skills/tree/HEAD/packs/ai-for-customer-service-pack
- Page: https://pluginboard.ai/plugins/ai-for-customer-service-pack

## Works with

- Claude
- GitHub Copilot
- Claude Code
- OpenClaw

## Install

### Claude

In Claude (web or desktop): Customize → Plugins → Add → Add marketplace, enter polar-bear-org/claude-skills, then add ai-for-customer-service-pack

### Claude Code

In Claude Code:

```text
/plugin marketplace add polar-bear-org/claude-skills
/plugin install ai-for-customer-service-pack@polar-bear-skills
```

---
From Pluginboard (https://pluginboard.ai).