Third-party integrations
AlignAI is designed as a connective layer that sits on top of the tools your AI program already runs on — it pulls work and data in from intake, GRC, delivery, and reporting systems, applies a standardized governance and risk workflow, and pushes outcomes back out, rather than replacing any of those tools.
USAA integration plan
This is the specific set of systems scoped for this engagement, as provided by USAA. It replaces a general connector-availability list with what’s actually being built for this deployment. Where a system needs a named contact on USAA’s side to move forward, that’s noted in the last column.
| USAA system | Integration description | Setup | Owner (your side) |
|---|---|---|---|
| Eagle GPT (Gemini) | Share the central “front door” and push intake to AlignAI to refine and approve (through API or MCP). | One API/MCP call on submit to create the use case. AlignAI supplies the contract and a worked example. | Eagle GPT team |
| Jira | Create, link, and sync Jira Epics or Tasks to manage delivery and view telemetry downstream. | Project plus service account, create/read scope. | PMO / Jira admin |
| ServiceNow | Create, link, and sync ITSM records (CMDB) for tracking. | Scoped table access plus integration user. | ServiceNow platform team |
| Homegrown GRC / MRM | Pull from policies to manage risk criteria (in the Context Layer) and push risk reviews for regulation/audit. | Target schema plus an endpoint, or an agreed scheduled export. One 60-minute working session first. | GRC owner |
| Arize / Dynatrace | Pull telemetry (traces) and usage metrics. | Read access to project and usage APIs; provider admin usage APIs. | Architecture, FinOps |
| Databricks | Pull datasets and metadata from Unity Catalog to provide data-set linkage. | Service principal, Unity Catalog read. Confirm the catalog of record. | Data platform |
| GitHub | Pull context to populate Blueprint and records in AlignAI from project artifacts and performance information. | GitHub App or token on the Enterprise instance, plus network path. | Platform engineering |
| AWS Guardrails | Define rules for implementation in AWS Guardrails through the codebase. | Managed through the GitHub integration / AlignAI MCP. | Platform engineering |
Eagle GPT and Homegrown GRC/MRM are custom to USAA’s environment rather than pre-built connectors; their setup follows the general pattern below (use case, field mapping, technical contact, auth).
Design principles
- No new logins. Integrations are meant to reduce systems, not add them.
- Bidirectional where possible. Two-way sync is the target architecture for every integration; Jira is bidirectional today, others are unidirectional while full sync is built out.
- Least-privilege by default. Every integration is scoped to the minimum permissions needed for its specific data flow — AlignAI does not aggregate access into one broad service account.
- Authentication depends on the system, not a fixed AlignAI requirement. Setup uses Basic Auth, OAuth, or a service account, depending on what each system supports — and the specific credential (an API token, a bearer token, a client ID, etc.) is whatever that system requires, not something AlignAI mandates. A few systems need an extra identifier (Jira, for example, uses a Cloud ID when an instance has multiple projects); your AlignAI contact will guide you to the correct values for your specific setup.
What we need from you to set one up
Describe the use case
What problem you’re solving and which direction data should flow.
Map your data fields
This is the primary driver of setup complexity, so get it nailed down early. See Field configuration & data mapping.
Name a technical point of contact
The “owner” column above is the starting list for this engagement.
Provide authentication access
An AlignAI representative will guide your team to whatever that specific system requires. You don’t need to hand credentials directly to the AlignAI team.
The most common source of delay is waiting on internal IT queues for service-account approval or network/firewall rules — worth starting those requests for each owner above in parallel, rather than sequentially.
Testing before production
Sandbox testing is supported and encouraged for every integration. For the custom integrations (Eagle GPT, Homegrown GRC/MRM), expect the working session noted in the table above to double as the first sandbox test.