Mapping ReBAC Across
Two Merged ERPs
When two manufacturers merge their ERP estates, access control work usually starts with a role mapping table. That artifact is wrong, and wrong in a way that can stay hidden for a year. What you are mapping is not roles. It is relations.
Table of Contents▼
- › 2.1 Pull the Relation Vocabulary Before the Role List
- › Relations Settle After Object Types
- › Relations That Cross the Entity Boundary
- › Tracking Relationships Over Time
- › Budget for Evaluation Depth
- › 2.2 AI Makes the Authorization Model More Important
- › 2.3 The Warehouse Is the Last Part Left
- › FAQs
When two manufacturers merge their ERP estates, the access control work almost always starts with a role mapping table. Someone exports the role lists from both systems, places them side by side, and starts drawing lines between them - Sales Manager East on the left, Regional Sales Director on the right.
That artifact is wrong, and wrong in a way that can stay hidden for a year.
A role in an ERP is not just a set of permissions. It is a permission set plus an implicit data scope, and the scope often carries most of the meaning. The Plant Controller for Plant 3 and the Cost Accountant can have access to the same transaction codes and still be authorized to do entirely different things because one is scoped to a plant and the other to a cost center.
| Role | Transaction Codes | Data Scope | What They Can Actually Do |
|---|---|---|---|
| Plant Controller (Plant 3) | Same | Plant | Act on data within Plant 3 |
| Cost Accountant | Same | Cost center | Act on data within their cost center |
What you are mapping is not roles. It is relations.
2.1 Pull the Relation Vocabulary Before the Role List
Review both ERP environments and identify the relationships that actually determine access. In manufacturing, these typically include:
Account ownership
A user owns an account.
Territory management
A user manages a territory containing specific accounts.
Plant membership
A user belongs to a plant.
Reporting line
A user supervises an employee.
Approval authority
A user has approval authority above a defined threshold for an entity.
Product line assignment
A user is assigned to a product line that contains specific items.
Both companies will have many of these relationships. They will not necessarily implement them the same way.
| System | How Sales Visibility Is Controlled | Access Is Granted |
|---|---|---|
| Legacy A | By territory | Based on territory membership |
| Legacy B | Specific accounts assigned to individual users | Account by account |
After the merger, you need to decide which approach becomes the standard, or maintain both and deal with two separate authorization paths.
That choice should be an architectural one, not one dictated by whichever migration script is written first.
Relations Settle After Object Types
Relations point to objects, so you must resolve the object model first: account, order, plant, cost center, price list, item, employee record, warehouse, and legal entity. They may all exist in both ERP estates, but under different names and at different levels of granularity.
Customer granularity
One system may authorize at the customer-parent level while another works at individual sold-to accounts.
Warehouse boundaries
One may consider warehouse locations part of a plant, while another treats them independently.
Until these object boundaries are resolved, the relationship model cannot be trusted.
It is essentially the item master problem applied to authorization: upstream of everything and often discovered after downstream work has already started.
Figure Out the Relations That Cross the Entity Boundary
This is where the merger-specific complication begins. Before the merger, most relationships existed within a single entity. Now a rep from legacy A can sell a legacy B product line into an account previously owned by legacy B, with fulfillment coming from a plant inherited from A.
The relation "owns account" now requires more context. Whether a user can see margin on an order may depend on:
- which entity booked the order,
- which entity manufactured it,
- which product line is involved, and
- when ownership of the account changed.
Entity context therefore needs to be part of the authorization model, not an assumption hidden inside a role.
Decide Whether Relationships Need to Be Tracked Over Time
Access control is not only about who can see something today; during an audit, you may need to show who had access at a specific point in the past, such as March.
The relationship model most teams use, based on tuples of object, relation, and subject, is essentially a current-state model. It efficiently answers authorization questions about now. Historical reconstruction requires versioned relationships or another authoritative history store.
If territory ownership shifts on April 1, simply replacing the relationship may remove the previous assignment from the current authorization state. Supporting historical access means preserving effective dates and, ideally, the policy version that interpreted those relationships at the time.
Deciding not to support historical evaluation is defensible. Not making the decision is what creates the problem.
The first post-consolidation audit can produce a question the system structurally cannot answer, leaving the team to reconstruct access manually from change logs and migration records.
Budget for Evaluation Depth Before Designing Relationship Chains
In manufacturing, access decisions can span multiple connected objects. A user may be linked to a plant, cost center, work order, item, and standard cost. As those chains grow, each additional relationship can add lookup and evaluation overhead, making performance an important consideration when designing the authorization model.
If someone opens one order, a few milliseconds for an authorization check does not matter. The same approach becomes expensive when a consolidated margin report returns forty thousand rows.
This creates an important technical distinction: point authorization and list authorization are two different workloads.
| Workload | Example Question | How It Is Handled |
|---|---|---|
| Point authorization | "Can this user view order 12345?" | Evaluated through the relationship graph |
| List authorization | "Return every order this user can view" | Generally cannot be implemented by running the same check tens of thousands of times. Usually requires reverse indexes, materialized permissions, authorization-aware query filters, or cached access sets |
That introduces another problem: invalidation. If a salesperson's territory changes on the first of the month, all derived account permissions based on that relationship may change as well.
The system needs to propagate the change quickly enough that cached authorization does not continue granting stale access.
2.2 AI Makes the Authorization Model More Important, Not Less
Once an AI assistant can query orders, summarize customer margins, create transactions, or operate across ERP and WMS systems, authorization can no longer stop at the application screen. The AI needs to inherit the authorization context, relationships, and constraints of the user or process it represents.
A sales copilot asked to "show me our highest-margin accounts" should not simply retrieve every account the underlying model or application can technically reach. The request still has to respect:
Account access
Which accounts the user is allowed to access.
Legal entity
Which legal entity booked the revenue.
Product line scope
Which product lines are within scope.
Margin visibility
Whether that user is permitted to see margin data.
Agentic workflows make this more important because one task can cross several systems. An AI agent might:
Start in CRM
The task begins with customer and account context.
Retrieve pricing from ERP
The agent crosses into the ERP for pricing data.
Check inventory in WMS
The agent queries the warehouse management system.
Initiate an approval workflow
The agent triggers an action on the user's behalf.
The relevant authorization context needs to survive that chain rather than being lost when the agent crosses an application boundary.
2.3 The Warehouse Is the Last Part Left
WMS and execution systems have their own access models, often cruder than the ERP model: badge, terminal role, work area, supervisor, shift, or equipment qualification.
But authorization in a warehouse depends on relationships that often live in HR: who reports to whom, who is qualified on which equipment, who belongs to which facility, and who is currently scheduled on a shift.
Post-merger, there may be two WMS platforms and two HR systems. A relation such as "supervises" has to resolve consistently across all four before it can safely authorize a labor transaction or operational override.
Qualification also introduces time into the model. An individual may be qualified to operate equipment today and not next month. These operational relationships can therefore require effective dates and expiration rather than simple present-or-absent membership.
Why the Role Mapping Table Belongs Later
This is why the role mapping table belongs later in the consolidation:
Define the objects
Resolve the object model and its boundaries first.
Define the relationships
Identify the relationships between those objects.
Identify system owners
Determine which systems own that data.
Work through cross-entity scenarios
Bring entity context into the model.
Decide on historical access
Choose how historical access should be handled.
Account for depth and bulk queries
Plan for evaluation depth and list authorization.
Map the legacy roles
Only then map legacy roles into the new structure.
At that point, roles become useful again as a projection of a relationship model whose semantics are already understood.
We are mapping how this is being handled across specialty vehicle and safety equipment manufacturing, a segment that has changed ownership faster in the last eighteen months than in the previous ten years.
The important question is not how two role catalogs get combined. It is whether the authorization model underneath them can represent the relationships the combined manufacturer actually operates on.
And increasingly, that question extends beyond employees. ERP consolidation is no longer only about whether people can access the correct data. The authorization architecture being built today will also determine whether future AI assistants and agents can safely operate across the combined enterprise.
FAQs
Get Your Authorization Model Right Before You Merge Roles
Set up a consultation with GrayCyan to map the objects, relationships and entity context behind your ERP consolidation.
Schedule a ConsultationLooking for AI advice at your company? Talk to our Editor-in-Chief
As featured in
Unlock the Future of AI -
Free Download Inside.
Get instant access to HonestAI Magazine, packed with real-world insights, expert breakdowns, and actionable strategies to help you stay ahead in the AI revolution.