Customer Onboarding Checklist — Agent Intake v1.0.0-preview¶
A single-file checklist for customer admins deploying Agent Intake into a pilot tenant. Pair this with pilot-deployment-runbook.md for detailed steps.
Status note: This is a preview release. Do not deploy broadly until pilot-firm walkthrough and your firm's AI Governance Committee + InfoSec + Compliance + Legal + IT review have signed off.
Stage 0 — Prerequisites¶
Licensing and tenant¶
- Microsoft 365 E3 or E5 with Power Platform licensing for makers
- Power Pages license for the portal
- Microsoft Entra Agent ID available in the target tenant/cloud
- Microsoft Purview Records Management enabled for retention labels
- Microsoft Teams enabled for sponsor adaptive cards
Identities and permissions¶
- Power Platform Admin account available
- Microsoft 365 Records Management Admin account available (one-time retention label setup)
- Microsoft Entra Agent ID Administrator, Cloud Application Administrator, or Global Administrator available for Agent ID setup/consent
- Automation identity selected (managed identity/workload identity for automation; interactive admin accepted for pilot setup)
- Microsoft Graph permissions reviewed:
-
User.Readdelegated for maker pre-fill;User.Read.Allonly if admin-scale lookup is needed - Delegated
RecordsManagement.Read.AllorRecordsManagement.ReadWrite.Allfor Graph beta retention-label verification/create, or manual Purview verification instead -
AgentIdentity.CreateAsManagerorAgentIdentity.Create.Allfor Agent ID creation - Agent Identity blueprint created or identified; ID stored for
AGENT_INTAKE_AGENT_BLUEPRINT_ID
Repo and tooling¶
- Python 3.10+ installed
- PowerShell 7+ installed
- Python dependencies installed (
requests,msal,pyyaml,azure-identityas needed) - Repo cloned:
git clone https://github.com/judeper/FSI-AgentGov-Solutions.git - You can run
python agent-intake/scripts/seed_classification_rules.py --self-testand see all PASS
Governance sign-off¶
- Pilot scope agreed (department, makers, duration)
- Sponsor list identified
- InfoSec aware of 10% sample-audit cadence
- Compliance aware of FINRA 3110 attestation language in
templates/sponsor-approval-card.json - Legal/Records aware of FINRA 4511 / SEC 17a-4 / CFTC 1.31 retention assumptions
- Records Management aware of the
FSI-AgentIntake-7yrPurview label
Stage 1 — Customize policy defaults¶
- Review
docs/decisions.md - Update
templates/policy-lookup-tables.yamlfor sponsor SLA, sample rate, data residency, and retention labels - Review
templates/sponsor-approval-card.jsonattestation language with Counsel
Stage 2 — Deploy schema and labels¶
- Dry-run:
python scripts/create_fsi_intake_dataverse_schema.py --dry-run --environment-url <pilot-env-url> --auth-mode managed-identity - Apply:
python scripts/create_fsi_intake_dataverse_schema.py --environment-url <pilot-env-url> --auth-mode managed-identity - Regenerate docs:
python scripts/create_fsi_intake_dataverse_schema.py --output-docs docs/dataverse-schema.md - Records Admin: run
python scripts/setup_purview_retention_label.py --output ./.agent-intake-smoke/label-spec.json - Create/verify
FSI-AgentIntake-7yrin the Purview portal or via Security & Compliance PowerShell - Verify label if delegated Graph beta access is approved:
python scripts/autodetect_purview.py --label-name FSI-AgentIntake-7yr --token-source cli
Stage 3 — Wire identities and consent¶
- Run
python scripts/setup_entra_agent_id.py --check-consent --token-source cli - Confirm
AgentIdentity.CreateAsManagerorAgentIdentity.Create.Allis consented for the handoff identity - Dry-run handoff with
--blueprint-id <agentIdentityBlueprintId>before enabling Flow 3 - Verify environment list:
python scripts/autodetect_environments.py --output ./.agent-intake-smoke/environments.json --token-source cli
Stage 4 — Build maker surface¶
- Power Pages site provisioned
- Follow
portal-configuration.md - Form binds to
fsi_intakerequest - Trigger questions T1–T6 use Yes / No / Not sure text values
- Sponsor field auto-populates from Graph
/me/managerwith override allowed - Test as a non-admin user
Stage 5 — Build workflow¶
Per flow-configuration.md. All flows are built manually in Power Automate designer.
- Flow 1 — classifier/router
- Flow 2 — Teams sponsor card using
Action.Submit - Flow 3 — handoff and decision evidence
- Connections and environment variables documented in customer-side ALM
Stage 6 — Validate end-to-end¶
- Run
pwsh agent-intake/scripts/smoke_test.ps1 - Submit a test intake as a known maker
- Verify auto-classification is Tier 3 / Zone 3 / Express
- Verify sponsor receives Teams card
- Sponsor approves
- Verify Microsoft Entra Agent ID service principal is created
- Verify entry created in
agent-registry-automation - Verify decision-log row has
FSI-AgentIntake-7yrlabel evidence
Stage 7 — Pilot kickoff¶
- Distribute
maker-quick-start.md - Distribute
sponsor-cheat-sheet.md - Schedule 30-day InfoSec sample review
- Schedule 90-day value review with the AI Governance Committee
- Set up monitoring per
drift-detection-integration.md
Stage 8 — Go / no-go decision¶
Before scaling beyond the pilot cohort, confirm:
- No undetected high-risk requests passed through Express
- Sponsors respond within SLA on average
- No regulatory or supervisory finding raised by Compliance against an approved agent
- Maker satisfaction informally positive
- AI Governance Committee approves the next cohort
Rollback¶
- Disable the three Power Automate flows
- Disable or hide the Power Pages portal page
- Preserve existing Dataverse decision-log rows
- Document rollback in customer change management