Case studies · Global capability center, India

A global capability center in India runs AI under its US parent's policy

Approved AI for the center's engineers and analysts, with the US parent's policy applied to each call, personal data redacted, cost charged back by business unit, and one record the parent's security team can read.

TowerAI spendCharged back by business unit
AI spend by business unit served

$0

  • Cards0%
  • US Retail0%
  • Wealth0%
  • Shared services0%
Alert sent to the Cards delivery lead: 84% of the monthly budget. Work carries on.
Illustrative product view. Sample data, not client data.

The problem

The center does engineering, analytics and operations for its US parent. Its people want AI tools. The parent's security team wants the same controls in India as at home, the center also has India's DPDP Act to meet, and finance wants AI cost charged to the business unit each team serves.

  • Onepolicy across the US parent and the India center
  • Per unitAI cost charged back to the business served
  • One recordthe parent's security team can read

Before Raidu

  • Different AI rules in each location
  • Work data pasted into personal tools
  • AI cost in one shared bill

With Raidu

  • The parent's policy applied in India too
  • Personal data redacted before the model
  • Cost by business unit, with alerts

The challenge

A global capability center works on its parent’s systems and data, so the parent’s security rules travel with the work. The center’s teams were already using AI, some on personal accounts, and the parent had no view of what they shared. The center is also a data fiduciary under the DPDP Act for the personal data it handles.

How they set it up

Raidu runs in the parent’s own AWS account. The center’s people use one workspace with the models the parent approves, and the parent’s policy runs before each call: personal data is redacted, and coding agents have keys and tokens blocked before a prompt leaves. Each team has a budget tagged to the business unit it serves, with alerts.

What changes

The parent’s security team reads one record for the US and India alike. Finance charges AI cost to the business that used it. The center’s teams keep working with the tools they asked for.

Rollout

  1. Step 1Raidu installed in the parent's own AWS account
  2. Step 2Workspace opened to a first delivery team
  3. Step 3Coding agents and budgets per business unit added

What they deployed

  • AI Console: one workspace with the models the parent approves
  • AI Firewall: the parent's policy applied, and personal data redacted before the model
  • AI Spend: a budget per business unit served, with alerts
  • Coding Agents: secrets blocked in Cursor and Windsurf

Talk to us about your data and AI.

Tell us what you need to prove and to whom. We will show you the record it produces.

Talk to us