← All case studies
Case study · Global bank, Americas · Model risk · delivered via partner

Model risk inventory moved off Archer onto Power Platform

How a model risk management function replaced its legacy GRC inventory with a governed Power Apps system of record inside the bank’s own Microsoft tenant, with committee reporting on the same model.

Client
Americas arm of a global bank
Function
Model Risk Management
Stack
Power Apps · Dataverse · Power BI · Power Automate
Period
[START] – [END], [X] people

Context

The model risk function kept its inventory, validation findings and attestations in a legacy GRC platform whose licence, administration and change cost had outgrown what the team used it for. Reporting to the model risk committee was assembled separately, by hand, and did not always agree with the inventory. [CLIENT-SPECIFIC CONTEXT — with permission]

Approach

We prototyped the inventory module on our own tenant against anonymised data, so the function could see the process running before any client system was touched. The production build then ran inside the bank’s Microsoft tenant, on Dataverse, with the legacy platform kept live until each module was retired on its own evidence. Reporting was built on the same model from day one, so the committee pack and the inventory could not disagree.

What was built

Module
Scope
Model register
Inventory with ownership, tiering, lifecycle status, dependencies and change history.
Validation and findings
Validation scheduling, findings with severity and owners, remediation workflow, evidence attachments.
Attestation and approvals
Periodic attestations by model owners, approval routing, audit trail of who approved what and when.
Committee reporting
Power BI on the Dataverse model: inventory health, overdue validations, findings ageing, tier concentration.

Outcome

[X]models migrated
[X wks]kick-off to first module live
[X%]of prior annual platform cost

[CLIENT QUOTE OR SPONSOR STATEMENT — with permission]

What we would bring to the next one

The register, findings and attestation modules are now accelerators. A second institution starts from a working system on anonymised data and spends its budget on its own taxonomy and integrations, not on the plumbing.

Running a model inventory on a licence you would rather not renew?

We will show you the register module on anonymised data in one call.