AI in Resolve
Resolve uses AI in two distinct ways, and it is worth separating them because the risks differ.
| Configuration generators | Operational AI | |
|---|---|---|
| Does | Drafts settings for you to review | Acts on live tickets |
| Runs | When you press a button | Continuously |
| Risk | A bad draft you can edit | A wrong answer reaching a customer |
| Where | Five settings screens | Runbooks, deflection, the assistant |
1. The five configuration generators
Each drafts a starting configuration you then edit. Nothing is saved until you accept it.
| Generator | Where | Permission |
|---|---|---|
| Generate Workflow | Statuses & Workflows | Create Service Desk Status and Workflow with AI |
| Generate Custom Fields | Custom Fields | Create Service Desk Custom Fields with AI |
| Generate Service Catalog | Service Catalog | Create Service Desk Catalogue with AI |
| Generate Classification | Ticket Classification | Create classification with AI |
| Generate Release Stages | Release Board | Create Service Desk Release flow with AI |
Their value is escaping the empty screen. On a project already carrying tickets, a generated workflow or taxonomy replaces structure that live records depend on. Generate early; edit by hand later.
It produces a plausible, conventional configuration. Plausible is not correct — it cannot know that your Pending status must pause the SLA clock, or that a laptop request needs three approvals. Review every field, especially the Pause SLA flags.
2. Operational AI
| Feature | What it does |
|---|---|
| AI Runbooks | Standing diagnostic and action steps per intent |
| AI training sources | The corpus the assistant answers from |
| Ticket deflection | Suggests articles while a customer writes a ticket |
| Deep Translate | Translates content, respecting the glossary |
Operational AI answers from what you ingested. An outdated document makes it confidently wrong — worse than having no answer, because it will be believed. Curate training sources and check chunk counts after ingestion.
3. Cost and governance
Every AI call consumes credits. AI Governance (/org/resolve/ai-governance)
reports daily calls, spend by model and by utility, tokens, and failures.
The governance view separates Calls, Credits, Tokens and Failed. A rising failure count with steady spend usually means a misconfigured utility retrying — worth checking before it becomes a bill.
4. Don't confuse these
| These two | Differ by |
|---|---|
| Generate with AI vs AI Runbooks | One-off configuration drafting vs standing operational instructions |
| AI training sources vs KB articles | What the AI reads vs what customers read |
| AI Governance vs Integration Telemetry | AI spend vs API traffic |