Customer-managed AI
Provider clarity without hidden model spend
Dchat lets teams separate the support workspace from AI provider usage so finance, security, and procurement can review model cost, key ownership, and data-flow decisions directly.
Bring provider ownership into the review
Customer-managed mode gives reviewers a cleaner ledger for who owns the AI key, where provider usage appears, and how model spend is approved.
- Key owner
- Usage ledger
- Finance approval
Keep prompt data categories explicit
Teams can document which conversation details, knowledge snippets, and tool context may be sent to the AI provider before launch.
- Conversation fields
- Knowledge snippets
- Tool context
Avoid per-resolution opacity
Separating platform cost from provider usage keeps pricing review clearer than bundling every model decision inside a black-box automation plan.
- Platform cost
- Model usage
- Review owner
AI provider decisions should be inspectable before launch.
| Capability | Status | Detail |
|---|---|---|
| Customer-managed provider mode | Available | Use customer-owned provider credentials when finance or security needs direct visibility into AI usage and data flow. |
| Dchat-managed provider mode | Available | Use Dchat-managed AI when a simpler launch is acceptable and signed customer terms cover provider responsibilities. |
| Prompt data review | Required | Document which customer fields, transcript snippets, knowledge articles, and tool context may enter prompts. |
| Usage and cost reconciliation | Configured | Reconcile Dchat platform usage, provider usage, finance approval, and buyer-facing claims before quoting savings. |
| Multi-model marketplace parity | Not claimed | Do not claim full model marketplace, automatic model comparison, or enterprise BYO-model hosting without shipped proof. |
| Provider-certified data boundary | Evidence required | Provider certifications, data-retention guarantees, and regional processing promises must come from signed provider evidence. |
This closes the cost and governance gap by making model usage a reviewable decision, not a hidden ingredient inside automation pricing.