Exploring the AI Helpdesk API with AI

​

AI Integration

• Natural-language data lookups (“show me unhandled sessions”)
• Trend analysis and summary reports
• AI Agent workflows that reference knowledge articles and procedures

​

Resources for AI Developers

​

Using the OpenAPI Schema

https://assist.i.moneyforward.com/api/v1/openapi.json

​

Building a Natural-Language Interface

• “List 10 unhandled sessions”
• “Group sessions closed in the last 7 days by assignee type”
• “Show all knowledge articles in the onboarding category”

​

Code Generation

1. Describe what you want in plain language
2. The AI picks the right endpoint (e.g. /api/v1/sessions, /api/v1/knowledge)
3. Get runnable code with the X-API-Key header set

​

Best Practices

​

Security

• Never hand the AI tool your API key in plain text — load it from an environment variable
• Review AI-generated code for security issues before running it
• Avoid issuing over-privileged keys (only grant ADMIN role when needed)

​

Error Handling

• Have AI-generated code branch on {"ok": false, "error": ...} rather than HTTP status alone
• Keep request rate below ~20 RPS per key to avoid upstream throttling

​

Data Privacy

• Mask session content and personally identifiable information before feeding it to the AI
• Review the AI tool’s data retention policy before sending sensitive content

​

Sample Use Cases

​

Daily Report Generation

# Prompt: "Summarize open and in-progress sessions by assignee type"
# 1. GET /api/v1/sessions/stats — counts by status
# 2. GET /api/v1/sessions?status=open — the list
# 3. AI writes a narrative summary

​

Knowledge Coverage Analysis

# Prompt: "Suggest knowledge articles that should exist based on unresolved sessions"
# 1. GET /api/v1/sessions?status=open
# 2. GET /api/v1/knowledge
# 3. AI identifies the gap

​

Support