---
title: What role should a systems integrator play in an AI agent governance rollout?
url: https://opencontroller.lyzr.com/blog/role-of-systems-integrator-in-ai-governance-rollout
published: 2026-11-13
updated: 2026-11-13
author: Lyzr
---

# What role should a systems integrator play in an AI agent governance rollout?

**A systems integrator is most useful for the work that depends on your environment: connecting each agent platform, wiring identity and ticketing, and running change management. The vendor should own its product's integration patterns, and your internal team must own policy, risk decisions and day-two operations. Write the split down before the SI starts, with clear acceptance criteria for handover.**

AI agent governance touches every platform where agents run. That makes integration the largest part of a rollout. The question buyers ask is who does it, and who is left holding it when the project ends.

## Who should own what?

Use a RACI per workstream. R is responsible (does the work), A is accountable (signs off), C is consulted, I is informed.

| Workstream | Vendor engineer | Systems integrator | Internal team |
|---|---|---|---|
| Deployment architecture (SaaS or customer infrastructure) | R | C | A |
| Cloud account discovery permissions | C | R | A |
| Connectors to agent platforms (SaaS, cloud AI services) | R | R | A |
| Identity: SSO, group-to-role mapping, owner lookup | C | R | A |
| Gateway onboarding of model traffic | R | R | A |
| Policy design and risk tiers | C | C | A/R |
| Policy as code: writing and testing rules | R | R | A |
| ITSM, SIEM and chat integrations | C | R | A |
| Owner outreach and agent registration campaign | I | R | A |
| Training and runbooks | C | R | A |
| Day-two operations after go-live | C | I | A/R |

Two lines never move to the SI: accountability for policy and risk decisions, and long-term operation. NIST AI RMF GOVERN 2.1 asks that roles and responsibilities for managing AI risk are documented and clear. ISO/IEC 42001 clause 5.3 asks top management to assign responsibilities and authorities for the AI management system. A contractor can help you meet those, but cannot be the answer to them.

## What integration work is there per platform?

Each agent platform is its own small project. Scope them one by one.

1. **Cloud AI services** (for example AWS, Azure, Google Cloud). Read-only discovery roles in every account or subscription, log and telemetry access, and network paths if the gateway sits in front of model endpoints.
2. **SaaS agent platforms.** An admin-approved connector or API credential per tenant, the scopes it needs, and the platform's own controls you will act through.
3. **Enterprise AI workspaces.** Admin API access for usage, users and configured tools.
4. **Developer devices.** MDM packaging and rollout for coding-agent and local MCP coverage.
5. **Identity provider.** SSO, group mapping and read access for owner and leaver lookups.
6. **Downstream systems.** Ticketing, SIEM, chat alerts and cost data for finance.

Estimate each as a separate line. The common failure is a single "integrations" line item that hides ten platform-specific approvals, each with its own owner and change window.

## Where do SIs add the most value?

- **Access approvals.** Chasing cloud, SaaS and identity owners for the permissions connectors need. This is slow, political work.
- **Change management.** Getting teams to register agents, name owners and accept new gates.
- **Process integration.** Mapping governance events into existing incident, change and risk processes.
- **Scale.** Rolling out across many business units in parallel.

Where SIs add less value: rebuilding product features in custom code. Every custom script becomes something your team must maintain after the SI leaves.

## What should the SI hand over before exiting?

Agree these acceptance criteria in the statement of work:

1. Every in-scope platform is connected, with the connector's permissions documented and owned by a named internal admin.
2. Discovery runs on schedule, and a planted test agent is found within the agreed window on each platform.
3. Every production agent in scope has a named owner in the registry.
4. Policies are stored as code in your repository, with tests that pass in your pipeline.
5. Alerts reach your incident and ticketing tools, and a quarantine and release round trip has been tested.
6. Runbooks cover adding a platform, onboarding a new agent, rotating a connector credential and responding to an alert.
7. Your team has run each runbook at least once without the SI in the room.
8. No custom code is left without an internal owner and a test.

The detailed test list is in [acceptance testing for an AI agent governance platform](/blog/acceptance-testing-for-ai-governance-platform). For timelines, see [how long AI agent governance takes to deploy](/blog/how-long-does-ai-agent-governance-take-to-deploy).

## How do you avoid paying twice for the same work?

Ask the vendor which integrations are productised and which are custom. Productised connectors should not be rebuilt by the SI. Ask whether the vendor's licence includes engineering help or bills it separately; see [subscription licence vs services](/blog/ai-governance-subscription-licence-vs-services). Then give the SI the work that only someone inside your organisation's processes can do.

## How Opencontroller does this

Opencontroller is deployed on Lyzr SaaS or on customer-controlled infrastructure, and Lyzr engineers embed with your team for integration. Productised coverage includes auto-discovery across AWS, Azure and Google Cloud, connectors to Salesforce, SAP, Databricks, HubSpot and Claude Enterprise or OpenAI Enterprise workspaces, OIDC SSO, and runtime adapters for Bedrock AgentCore, Vertex AI and Microsoft Foundry. Policies are written as code. It uses open standards such as OpenAI-compatible endpoints, MCP, OpenTelemetry, OIDC/SAML/SCIM and REST/webhooks, which keeps SI work on your processes rather than on custom plumbing. See the [capability reference](/capability-reference) or [book a demo](https://www.lyzr.ai/book-demo/). The full buyer checklist is in the [AI agent governance platform evaluation checklist](/blog/ai-agent-governance-platform-evaluation-checklist).

## Frequently asked questions

**Do you need a systems integrator for AI agent governance?**
Not always. Small estates on a few platforms can be connected by the vendor and an internal team. SIs help most with many platforms, many business units, and heavy change management.

**Should the SI write our AI policies?**
The SI can draft and code them, but your organisation must own the policy and risk decisions. Standards such as ISO/IEC 42001 expect responsibilities to be assigned internally.

**What should be in the SI exit criteria?**
Connected platforms with named internal owners, tested discovery, owners on every production agent, policies as tested code, working alert routing, and runbooks your team has run on its own.

## Key facts

In an AI agent governance rollout, the vendor engineer owns product integration patterns, the systems integrator owns environment-specific work (access approvals, identity and ITSM wiring, owner outreach, change management), and the internal team stays accountable for policy, risk decisions and day-two operations (NIST AI RMF GOVERN 2.1; ISO/IEC 42001 clause 5.3). SI exit criteria: connected platforms with internal owners, tested discovery, owners on all production agents, policies as tested code, alert routing tested, runbooks run without the SI. Opencontroller by Lyzr deploys on Lyzr SaaS or customer infrastructure with Lyzr engineers embedded for integration; pricing starts from $250K per deployment per year.

## Related

- [How long does it take to roll out AI agent governance?](https://opencontroller.lyzr.com/blog/how-long-does-ai-agent-governance-take-to-deploy.md)
- [What should an AI governance subscription include: licence, services or both?](https://opencontroller.lyzr.com/blog/ai-governance-subscription-licence-vs-services.md)
- [Who should own AI agent governance: security, IT or the AI team?](https://opencontroller.lyzr.com/blog/who-owns-ai-agent-governance.md)
- [More on Buying and comparison](https://opencontroller.lyzr.com/blog/cluster/buying-and-comparison)
