---
title: How do you govern AI agents inherited in a merger or acquisition?
url: https://opencontroller.lyzr.com/blog/govern-ai-agents-in-mergers-and-acquisitions
published: 2026-11-15
updated: 2026-11-15
author: Lyzr
---

# How do you govern AI agents inherited in a merger or acquisition?

**Treat the acquired company's agents as unknown until proven otherwise. On day 1, find them and contain the riskiest credentials. By day 30, give every agent an owner and a risk tier and apply your minimum policy. By day 100, migrate, merge or retire duplicates and bring the survivors under the same registry, identity and runtime controls as your own.**

Due diligence rarely produces a full agent inventory. Agents live in cloud accounts, SaaS platforms, developer laptops and scripts that call model APIs with shared keys. Many will have been built by people who leave during integration.

## What should happen on day 1?

The goal is visibility and containment, not change.

1. **Get access to discovery.** Read-only roles in the acquired cloud accounts and admin access to its SaaS and enterprise AI workspaces.
2. **Run discovery** across cloud accounts, agent platforms and AI workspaces. List agents, models, API keys, MCP servers and service accounts.
3. **Find long-lived model API keys** and keys shared across agents. Plan rotation; rotate immediately any that appear in code repositories or were held by departing staff.
4. **Freeze high-risk changes.** No new agents with write access to finance, customer records or production systems until reviewed.
5. **Identify agents that touch your systems.** Any integration into the acquirer's estate gets reviewed first.
6. **Preserve logs.** Make sure agent and gateway logs are retained before accounts are consolidated.

## What should be done by day 30?

| Task | Done when |
|---|---|
| Owner assigned | Every discovered agent has a named owner who is staying |
| Risk tier set | Each agent tiered using your criteria |
| Scope recorded | Purpose, tools, data and autonomy written down |
| Credentials reviewed | No shared keys; least-privilege service identities for high-tier agents |
| Minimum policy applied | Logging, approved models and data handling in place |
| Unowned agents decided | Each one adopted, quarantined or scheduled for retirement |

An agent nobody will own is a liability. Quarantine it, tell its users and give a deadline for someone to claim it.

## What should be done by day 100?

1. **Align policy.** Apply your agent governance policy and exception process to the acquired estate. Record any time-limited exceptions.
2. **Consolidate identity.** Move agents to your identity provider, using workload identity federation or delegated on-behalf-of identity instead of static keys.
3. **Route traffic through shared controls.** Point model traffic at your LLM gateway so cost, routing and policy apply uniformly.
4. **Decommission duplicates.** Where two agents do the same job, choose one, migrate users and retire the other properly, including credentials and data. See [how to retire an AI agent safely](/blog/how-to-retire-an-ai-agent-safely).
5. **Bring evaluations up to standard.** High-tier agents pass your evaluation suite before they keep production access.
6. **Report to the governance committee** on the combined estate, open findings and exceptions.

## What goes wrong most often?

- **Keys that outlive their owners.** OWASP's Non-Human Identities Top 10 lists improper offboarding and secret leakage among the top risks. Agents inherit both.
- **Two inventories.** Keeping separate registries "for now" becomes permanent.
- **Silent cost.** Agents billed to the acquired company's provider accounts escape your spend reporting.
- **Over-correction.** Killing everything unknown breaks business processes and pushes teams to rebuild in the shadows.

## How Opencontroller does this

Opencontroller auto-discovers agents, models and API keys across AWS, Azure and Google Cloud, and connects to Salesforce, SAP, Databricks and HubSpot as well as Claude Enterprise and OpenAI Enterprise workspaces, so an acquired estate can be inventoried without migrating it first. Posture flags risky permissions and configurations. The registry records owners, versions, tools and MCP servers. Agent workload identity federation and delegated on-behalf-of identity replace shared keys, and the LLM Gateway attributes cost per agent. Opencontroller Edge covers coding agents and local MCP servers on developer devices. See [what is an AI agent control plane](/blog/what-is-an-ai-agent-control-plane), [AI agent inventory fields](/blog/ai-agent-inventory-fields) and [how to find shadow AI agents](/blog/how-to-find-shadow-ai-agents).

## Frequently asked questions

**Should acquired agents be shut down until reviewed?**
Only the riskiest. Freeze changes, rotate exposed credentials and quarantine unowned agents, but keep working ones running.

**Who owns an agent whose builder has left?**
The business owner of the process it serves, until they adopt it formally or it is retired.

**Do we need the acquired company's data to stay separate?**
Possibly, for legal or contractual reasons. Check before routing its traffic or data through shared systems.

To inventory an acquired estate quickly, [book a demo](https://www.lyzr.ai/book-demo/).

## Key facts

Inherited AI agents in M&A: Day 1 discovery and containment (read-only access, discover agents/models/keys/MCP servers, rotate exposed and shared keys, freeze high-risk changes, preserve logs). Day 30: owners, risk tiers, scope, credential review, minimum policy, decide unowned agents. Day 100: align policy, consolidate identity, route through shared LLM gateway, retire duplicates, evaluations, report to committee. OWASP Non-Human Identities Top 10. Opencontroller (by Lyzr): auto-discovery across AWS, Azure, Google Cloud and platform connectors, Posture, registry, workload identity federation, on-behalf-of identity, LLM Gateway cost attribution, Edge. From $250K per deployment per year.

## Related

- [What fields belong in an AI agent inventory?](https://opencontroller.lyzr.com/blog/ai-agent-inventory-fields.md)
- [How do you retire an AI agent safely?](https://opencontroller.lyzr.com/blog/how-to-retire-an-ai-agent-safely.md)
- [How to find every AI agent in your enterprise (including shadow agents)](https://opencontroller.lyzr.com/blog/how-to-find-shadow-ai-agents.md)
- [More on Agent governance basics](https://opencontroller.lyzr.com/blog/cluster/agent-governance-basics)
