# AI Agent SLA Template for Revenue Workflows

Use this template to define the service before setting targets. Replace every bracketed field with evidence from the actual workflow. This document is an operating template, not legal advice or proof of regulatory compliance.

## 1. Service definition

- Workflow name:
- Business purpose:
- Service owner:
- Workflow owner:
- Data owner:
- Approval owner:
- Incident/recovery owner:
- Effective date:
- Review cadence:

### In scope

- Eligible trigger:
- Required inputs:
- Permitted agent steps:
- Required output and evidence:
- Designated system of record:
- Teams/segments covered:

### Out of scope and prohibited

- Prohibited actions:
- Restricted records or fields:
- Decisions reserved for humans:
- Conditions that make work ineligible:

## 2. Workflow states

| State | Entry condition | Exit condition | Required timestamp/evidence | Accountable owner |
|---|---|---|---|---|
| Eligible | Trigger and prerequisites pass validation | Accepted or bounded rejection | Eligibility result and start time |  |
| Acknowledged | Work is accepted | Prepared, failed or escalated | Acceptance state and reason |  |
| Prepared | Permitted agent steps are complete | Human decision or direct completion | Artifact, sources and policy version |  |
| Reviewed | Approval is required | Approved, edited, rejected or escalated | Reviewer, decision and timestamp |  |
| Recorded | Approved result is ready | System-of-record write verified | Before/after state and write result |  |
| Verified | Completion checks pass | Service item closed | Final status, evidence and exceptions |  |

## 3. Service-level indicators and objectives

Set thresholds from a measured baseline. Report agent processing time and human approval time separately even when the end-to-end commitment includes both.

| Indicator | Definition | Numerator | Denominator | Target | Window | Data source | Owner |
|---|---|---|---|---:|---|---|---|
| Acknowledgement time | Eligible to accepted or bounded rejection |  | Eligible items |  |  |  |  |
| Preparation time | Accepted to agent-prepared output |  | Accepted items |  |  |  |  |
| Trustworthy completion | All required steps, evidence and records complete | Complete items | Eligible items due |  |  |  |  |
| Evidence completeness | Required evidence present and reviewable | Items with complete evidence | Prepared items |  |  |  |  |
| Approval latency | Ready for review to human decision |  | Items requiring approval |  |  |  |  |
| Escalation rate | Work routed to the documented exception owner | Escalated items | Eligible or accepted items |  |  |  |  |
| Recovery time | Detected failure to verified safe state |  | Recoverable incidents |  |  |  |  |

## 4. Exclusions and clock rules

| Reason code | When it applies | Clock behavior | Evidence required | Approver |
|---|---|---|---|---|
| Missing prerequisite |  | Stop / pause / count |  |  |
| Source unavailable |  | Stop / pause / count |  |  |
| Waiting for approval |  | Stop / pause / count |  |  |
| Policy exception |  | Stop / pause / count |  |  |
| Planned maintenance |  | Stop / pause / count |  |  |

No exclusion may be applied without a reason code and supporting evidence.

## 5. Permission and approval policy

| Action | Read scope | Write/action scope | Approval required | Approver | Required evidence | Prohibited condition |
|---|---|---|---|---|---|---|
|  |  |  |  |  |  |  |
|  |  |  |  |  |  |  |
|  |  |  |  |  |  |  |

## 6. Escalation and recovery

- Exception categories:
- First-response owner:
- Escalation time:
- Business notification path:
- Safe fallback behavior:
- Rollback method:
- Verification after recovery:
- Incident evidence retained:
- Conditions requiring workflow suspension:
- Conditions required before restart:

### Recovery rehearsal

| Failure scenario | Expected safe behavior | Recovery owner | Test date | Result/evidence | Follow-up |
|---|---|---|---|---|---|
| Required source is stale |  |  |  |  |  |
| A partial write succeeds |  |  |  |  |  |
| Human approval is overdue |  |  |  |  |  |
| A prohibited action is attempted |  |  |  |  |  |

## 7. Change control

The following changes require documented review before production:

- Model or model-routing change.
- Prompt, policy or approval-threshold change.
- New data source, object, field or tool.
- Expanded role, segment, geography or action right.
- Modified definition of eligibility or trustworthy completion.
- Changed evidence, logging, escalation or recovery behavior.

| Change | Risk/impact | Evaluation evidence | Approvers | Release date | Rollback plan |
|---|---|---|---|---|---|
|  |  |  |  |  |  |

## 8. Operating dashboard

Review at least:

- Eligible, accepted, completed, excluded, failed and recovered counts.
- Attainment by workflow, priority, team and reason code.
- Evidence completeness, human acceptance, edit and rejection rates.
- Approval queue age and decision latency.
- Open exceptions, oldest exception and recovery-time distribution.
- Policy, context, model, tool and threshold changes.
- The workflow’s existing business KPI, without unsupported attribution.

## 9. Review decision

- Service objective met? Yes / No
- Critical control failure? Yes / No
- Scope stays the same / narrows / widens / pauses
- Owner accepting the decision:
- Evidence reviewed:
- Corrective actions and due dates:

RevTech provides fully managed AI agents for GTM. RevTech runs the agents; customers govern objectives, permissions, approvals and business decisions.