For most large enterprises, 2026 has brought two cost and capability levers to the same boardroom at the same time. The first is AI, moving from experimentation into production-scale operating models1; the second is the Global Capability Center, as enterprises continue to use owned global capability to build technology, operations and transformation capacity.
Both levers are sound. Both can support rigorous business cases. The problem arises when both cases are written for the same work by different teams, and approved months apart.
Savings double counting occurs when an AI business case and a GCC business case each claim savings against the same activities without accounting for each other. Each case is internally correct. Added together, they promise more than the underlying cost base can deliver, because the two levers act on the same work in sequence rather than independently.
Consider an illustrative finance and technology operations function with 100 roles and an annual cost of USD 20 Mn.
The GCC case estimates that performing this work from a capability center reduces cost by 60%, or USD 12 Mn. The AI case estimates that automation and augmentation remove USD 6 Mn of effort. Presented separately, a leadership team may reasonably conclude that the function will cost USD 2 Mn a year once both programs land.
The arithmetic does not hold. If AI removes USD 6 Mn of effort first, USD 14 Mn of work remains. Applying the same 60% location saving to that smaller base yields USD 8.4 Mn. The true combined saving is USD 14.4 Mn.
Exhibit 1. Separately built cases overstate the achievable integrated saving by 25%, (illustrative, USD Mn)
| Separate cases, added | Integrated case, sequenced | |
| Starting annual cost | 20.0 | 20.0 |
| AI savings | 6.0 | 6.0 |
| GCC savings (60% of remaining cost) | 12.0 | 8.4 |
| Total savings claimed | 18.0 | 14.4 |
| Resulting run-rate cost | 2.0 | 5.6 |
Source: Zinnov analysis; illustrative example.
The USD 3.6 Mn difference does not appear in either case. It surfaces in year two, when finance reconciles planned savings against actual run-rate cost.
Double counting is rarely a modeling error. It is the product of three structural conditions that are common in large enterprises.
Recent reporting and analysis suggests that AI investment is outpacing measurable returns: 60% of surveyed companies2 report minimal or no value from AI, including cost reductions or revenue gains. Meanwhile, finance leaders are contending with rising AI costs, complexity and the need for greater visibility3 into AI spending and performance. The risk of overlapping AI and GCC business cases is a separate but consequential issue that enterprises should test explicitly.
The financial shortfall is the visible cost. Three less visible costs tend to matter more.
High-performing GCCs are designed before they are staffed. The most reliable approach treats AI and location as parts of one decision and models them in a deliberate order.
Exhibit 2. Five steps to an integrated AI and GCC business case
| Step | Core question | Output |
| 1. Decompose the work | Which activities make up this function, independent of today’s roles? | Activity-level baseline of cost and volume |
| 2. Define the AI role | Which activities will be automated or augmented, and on what timeline? | AI impact curve by activity and year |
| 3. Define the human role | Where do judgment, accountability and customer context remain essential? | Target capability and talent profile |
| 4. Decide where work sits | Which remaining work belongs in the GCC, with a partner, or at headquarters? | Location and ownership model |
| 5. Build one model | What is the combined saving against a single cost base, in sequence? | Integrated case with dated assumptions |
Source: Zinnov GCC design methodology.
Two design choices make the integrated case durable. First, measure outcomes rather than roles transitioned: cycle time, cost per transaction or risk reduced. Second, give each major assumption a review date and an owner. A 5-year case built on today’s automation potential and talent costs will need re-underwriting well before year five.
Zinnov has designed, built and scaled more than 220 Global Capability Centers. Setting up a center quickly is valuable, and many enterprises need to move fast. The performance of that center, however, is largely determined by decisions made before launch: the business case, the work design, the role of AI and the decision rights the center will hold.
Zinnov’s Day-Zero design approach integrates AI and location strategy into a single business case, so the center is sized for the work the enterprise will need three years from now.
Not directly. Both levers act on the same work. AI savings should be applied first, and location savings calculated on the remaining cost base. Adding separately calculated savings typically overstates the total.
It depends on the function and the financial timeline. Some enterprises need near-term savings and move work before transforming it. The critical principle is to avoid underwriting multi-year savings against a workflow that AI is already changing, and to review those assumptions on a fixed schedule.
AI shifts the GCC cost model away from headcount-driven savings toward capability and productivity. Centers are increasingly sized by the outcomes and judgment-intensive work they own rather than by the number of roles transitioned.
A single executive owner, usually the CFO or COO, should own the combined savings number, with AI and GCC leaders accountable for their components against one shared cost baseline.
THANK YOU
FOR YOUR RESPONSE
