NirjiX
Where the two pillars meet

Design the center AI-native, not AI-retrofitted

Most capability centers add AI in year three, once the headcount model is locked and the roles are already wrong. Sequencing it first changes the size, the seniority and the economics of the center.

Six things that change when AI comes first

Capacity planning

Retrofit · Headcount is sized from today's process volumes, then AI is asked to justify itself against a headcount plan the board already approved.

AI-native · Capacity is planned on outcomes and cycle time. Automated and assisted work is modelled up front, so the center is smaller, more senior and cheaper to run at the same output.

Role design

Retrofit · Roles are ported from the parent org chart. AI arrives as a tool on top of unchanged jobs, so adoption depends on individual willingness.

AI-native · Roles are designed around review, exception handling and judgement. Fewer processing roles, more engineering and domain-owner roles, and a clear escalation path from model output to human decision.

Data and platform

Retrofit · Data access is arranged per project. Every AI use case restarts the legal, security and integration conversation.

AI-native · Data contracts, access tiers, residency and evaluation tooling are part of the entity and technology design, so the second use case takes weeks rather than quarters.

Governance

Retrofit · AI governance is added after an incident, usually as a review board that slows everything down.

AI-native · Model risk, evaluation thresholds, human-in-the-loop rules and audit trails are written into the center's operating charter before the first hire.

Business case

Retrofit · Savings are salary arbitrage. The board discovers ramp loss, attrition and management overhead in year two.

AI-native · The case carries two levers — location economics and automation yield — and states which benefits depend on AI delivery, so the downside scenario is honest.

Talent market

Retrofit · The center competes for the same generalist pool as every other captive in the city, on salary alone.

AI-native · The center offers AI-adjacent work that senior engineers actually want, which is the cheapest attrition control available.

The bridge path

The AI portal and the GCC portal answer different questions. Run them in this order and each one feeds the next.

  1. 01

    Establish AI readiness before you size the center

    Data maturity, platform state, governance and skills determine how much work can safely be automated or assisted. That number changes the headcount plan, so it belongs before the plan, not after.

    Run the AI assessment →
  2. 02

    Test whether the capability should sit in a center at all

    Some work travels well, some does not, and some should be automated instead of relocated. The feasibility assessment separates the three before anyone signs a lease.

    Run the GCC assessment →
  3. 03

    Model both levers in one business case

    Location economics and automation yield compound. Modelling them separately overstates one and hides the dependency between them.

    Build the business case →
  4. 04

    Design the operating model AI-native

    Role mix, data contracts, evaluation tooling and governance are cheap to write into the blueprint and expensive to retrofit into a running center.

    Generate the blueprint →
  5. 05

    Sequence delivery so the center proves value early

    A first wave that ships an AI-assisted process within two quarters buys the political capital for waves two and three.

    Build the AI plan →

Six questions a board should ask

  • Can we name the three processes where automation yield changes the headcount number by more than 15%?
  • Do we know which data cannot leave the parent jurisdiction, and what that removes from the center's scope?
  • Who signs off when a model output is wrong in production — and is that person in the center or at headquarters?
  • Does the business case state which savings disappear if AI delivery slips by a year?
  • Is there a single owner for evaluation and model performance, or is it split across vendors?
  • Are we hiring the seniority that AI-assisted work requires, or the volume that manual work required?
Start the GCC assessmentAssemble the investment pack

Frequently asked questions

What does an AI-native GCC mean in practice?
It means capacity is planned on outcomes rather than headcount, roles are designed around review and judgement instead of processing, and data contracts, evaluation tooling and model governance are written into the center's design before the first hire. The result is usually a smaller, more senior center with better unit economics than a like-for-like offshore team.
Should we run the AI assessment or the GCC assessment first?
The AI assessment first when the automation question is genuinely open, because how much work can be safely automated or assisted changes the headcount plan. The GCC assessment first when the mandate to build a center is already given and the question is model, location and sequencing.
Why is retrofitting AI into an existing center expensive?
Because the expensive decisions are already made: the role mix is hired, the data access is arranged per project, the governance is reactive, and the business case was approved on salary arbitrage alone. Changing any of those in a running center means renegotiating headcount, security posture and the board narrative at the same time.