# The AI support agent playbook for small teams

> An ungated playbook for teams of one to twenty: reduce support ticket volume with an AI agent in three parts (launch, evaluate, scale), with free CSV templates.

## The short answer

Put the answers to your repeat questions in sources an AI agent can read, let it resolve those intents end to end, and hand everything else to a person with context. Review weekly: every gap the agent reports is a source you write once.

## 1. Launch

- Tag the last 90 days of tickets by intent. If a few intents make up a big share, an AI agent can close them.
- Pick one goal (fewer tickets reaching a person, faster first replies, or after-hours coverage) and write it as a number.
- Record a baseline: tickets per month, top 20 intents, first response time, time to resolution, minutes per ticket, reopen rate.

## How an AI agent processes a support ticket

1. **Intake:** the message arrives from chat, email, or a form with customer context attached.
2. **Intent:** Kal reads what the customer wants, plus sentiment and urgency.
3. **Answer:** Kal answers only from the Knowledge Store (a repository of input sources such as help articles, PDFs, Notion, Drive) and cites the source.
4. **Handover:** when rules say a person should take it, Handover passes the thread with a summary of what was asked, tried, and read.
5. **Learn:** corrections stick, and missing answers become Knowledge Store gaps to fill.

## 2. Evaluate

Score any agent on: answer accuracy, citations, knowing when to stop, handover quality, setup time, pricing model, switching cost (Wire Through keeps live threads when you switch), and data handling. Build a 30 to 50 message test set from real tickets, score pass, partial, or fail, fix the source behind each fail, then go live on two or three intents.

## What 1,000 AI resolutions cost on SupportWire

$19 per seat per month, first 50 resolutions included, then $0.49 each. 1,000 resolutions cost $465.50 in AI fees at any team size; a 3-seat team pays $522.50 in total. Nothing is charged for a handover.

## 3. Scale

- Three hats, not three hires: an owner (about 2 hours a week), a knowledge keeper (about 1 hour a week), and everyone on the inbox correcting replies.
- Track AI resolution rate, handover rate, wrong answers, first response time, reopen rate, and hours given back at 30, 60, and 90 days.
- Assign the freed hours (missing sources, proactive onboarding outreach, fixing repeat bugs) before day 60, or the queue absorbs them.

## Templates

- [Handover rules template (CSV)](/templates/handover-rules-template.csv)
- [AI agent test set template (CSV)](/templates/ai-agent-test-set-template.csv)

See also: [Kal, the AI agent](/ai.md), [Pricing](/pricing.md), [Ticket deflection calculator](/free-tools/ticket-deflection-calculator)

---

Source: https://supportwire.ai/ai-support-agent-playbook
SupportWire. Resolution is the job.
Sign up: https://app.supportwire.ai/signup