Customer support · Production

Craft Docs builds Craft Agents — a visual Claude Code that non-engineers use more than devs

The problem

After three years of AI experimentation — including an early AI assistant, RAG-based knowledge search, and summarization features — Craft Docs had found nothing that retained users. A terminal-based prototype was also inaccessible to non-technical staff.

First attempt

The 2022 AI assistant gained an initial engagement spike but users did not return. Subsequent RAG and summarization experiments equally failed to achieve stickiness. A terminal-based Craft Agents prototype drew complaints from non-technical beta users about painful multitasking and awkward plan review.

Workflow diagram · grounded in source
1
Zendesk ticket arrives
trigger
“Look at an incoming Zendesk ticket”
2
Get User Data enrichment
integration
“It uses a Craft backend API to look up the user profile based on the user's email in the Zendesk field, and adds the user's plan type, billing status, feature flags (feature access), and usage metrics to the ticket”
3
Bug identification and tagging
ai_action
“Identify the platform and area which the bug belongs to, and tag it”
4
Linear crosscheck
integration
“Crosscheck with Linear (Craft's issue tracking system). If there are similar issues, link to that issue. If no similar issues are found, create a new issue and assign it to the relevant developer team”
5
Technical root cause analysis
ai_action
“Do a technical root cause analysis”
6
Response draft and ticket creation
output
“Draft a ready-to-send response for the customer. Create engineering tickets for developers, including pointing to code references, where applicable”
7
Parallel agent triage
ai_action
“When there's a large number of customer tickets, the tool automatically kicks off parallel agents to go through triage”
8
Triage report to support agent
output
“When triage is complete, the agent provides a report to the customer support person who kicked off the work”
Reported outcome

Craft Agents — built in two weeks on the Claude Agent SDK — became an instant internal hit.
Customer support ticket processing time fell from 20-30 minutes to 2-3 minutes, escalations to engineers dropped significantly, and non-engineering teams in marketing and HR began building automations without any developer input.

Reported metrics
Ticket processing time20-30 minutes to 2-3 minutes
Escalations to engineeringescalate so much less often than before
Ticket volume capacityprocess a much larger volume of tickets than before
Reported stack
Craft AgentsClaude CodeElectronZendeskLinearBamboo HRRevolutSlackRAG
Source
https://newsletter.pragmaticengineer.com/p/ai-first-makeover-craft
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Craft Agents — built in two weeks on the Claude Agent SDK — became an instant internal hit.

What tools did this team use?

Craft Agents, Claude Code, Electron, Zendesk, Linear, Bamboo HR, Revolut, Slack, RAG.

What results were reported?

Ticket processing time: 20-30 minutes to 2-3 minutes; Escalations to engineering: escalate so much less often than before; Ticket volume capacity: process a much larger volume of tickets than before (source-reported, not independently verified).

What failed first in this deployment?

The 2022 AI assistant gained an initial engagement spike but users did not return.

How is this customer support AI workflow structured?

Zendesk ticket arrives → Get User Data enrichment → Bug identification and tagging → Linear crosscheck → Technical root cause analysis → Response draft and ticket creation → Parallel agent triage → Triage report to support agent.

Related customer support cases