10decoders
The AI Center of Excellence Isn't a Decision-Making Body Yet
Walk into most GCCs today and you will find an AI Center of Excellence on the org chart. It has a name, a lead, a pilot inventory, and a slide deck for the quarterly business review. What it usually does not have is the authority to approve a use case, choose a vendor, or stop a pilot from shipping without checking with someone at headquarters first. Zinnov's analysis of 220-plus GCC engagements found that 92% of India's centers are piloting or scaling AI, yet more than 70% of their leaders report no structured way to measure whether any of it is actually delivering value. A center that cannot measure its own impact is rarely the one deciding what gets funded next.
For twenty years, the GCC model was built around one bet: that execution was the scarce resource. Give India five hundred engineers who can build what headquarters designs, at a fraction of the cost, and the center earns its keep. AI is breaking that bet, because the bottleneck is moving from capability to judgment, from what a team can build to what it should build next. Most AI Centers of Excellence were designed for the old bet. They track what is being piloted. They rarely decide what gets built.
The operating model that is supposed to fix this, a central hub owning platform and standards while individual business units run their own use cases underneath it, only works if the hub actually has authority over the decisions it is meant to own. Plenty of GCCs adopted the hub-and-spoke org chart in 2026 without transferring the decision rights that make it function. The shape looks right in a presentation. The authority is still sitting somewhere else.
Most AI Centers of Excellence were built to report on decisions somebody else already made.
Where GCC AI Decision Rights Actually Break Down
| Decision point | What usually happens instead | Severity |
|---|---|---|
| Use-case approval | Whichever business unit asks first gets funded, since no formal intake or prioritization step sits with the center | Critical |
| Vendor and model selection | Each pilot picks its own model or vendor locally, with no hub-level standard to consolidate spend or risk | High |
| Budget ownership | The center tracks AI spend but does not control a discretionary line, so every dollar routes through the annual headquarters cycle | Critical |
| Production sign-off | A pilot moves into production because it worked, not because anyone with real authority reviewed and approved the move | High |
| Policy enforcement | A governance policy exists on paper, but no one owns enforcing it against a business unit that ignores it | Moderate |
| Escalation to headquarters | Every non-trivial call still routes upward, even two or three years after the center was stood up | Moderate |
Not sure whether your GCC's AI center actually has decision authority?
10decoders reviews what your AI Center of Excellence is chartered to do against what it can actually approve, fund, or veto today, then builds a decision-rights matrix both the center and headquarters sign off on.
Book a Free AI Assessment →What Changes When the Center Gets Real Authority
When a GCC's AI Center of Excellence holds actual decision rights, the visible change is speed: a use case that used to wait three review cycles for a headquarters sign-off gets approved, funded, or rejected inside the center itself. The less visible change is accountability. A single team owns the call, which means a single team can be asked why a pilot got funded, why a vendor got picked, or why a use case never made it to production, instead of the answer scattering across five stakeholders who each remember the decision differently.
The Zinnov-Nasscom GCC Landscape report classifies India's 2,117 centers into four maturity tiers, and the split shows where most organizations actually sit. Thirteen percent remain Outpost centers, built for cost arbitrage. Forty-three percent, the largest single group, are Satellite centers delivering capability at scale, which is also the tier most enterprises get stuck in. Thirty-nine percent have reached Portfolio Hub status, with end-to-end ownership of a product or platform. Only 5% have reached Transformation Hub, where AI-led operations carry a real CXO mandate from India. Roughly 27% of centers reach Portfolio Hub within five years. The jump from Portfolio Hub to Transformation Hub is where decision rights, not headcount or capability, becomes the actual constraint.
The Center Tracks, Headquarters Decides
The center compiles a pilot inventory and presents it upward each quarter. Every material call, budget, vendor, production sign-off, still gets made outside India.
The Center Recommends, Headquarters Confirms
Technical standards start forming, a preferred model list, a security baseline, but every recommendation still needs a separate approval step before anything ships.
The Center Approves, Vetoes, and Funds
The center holds a real budget line, can say no to a use case without escalating, and owns production sign-off for the AI systems inside its charter.
Does Your Charter Actually Answer These Questions?
Most AI Center of Excellence charters read well and answer almost nothing. Test yours against the questions that actually determine whether the center can act or only advise.
AI Decision-Rights Checklist
The GCCs pulling ahead this year gave their AI center a budget line before they gave it a slide deck.
What to Do This Week
01 Pull last quarter's AI decisions and trace who actually made each one
List every AI use-case approval, vendor selection, and budget request from the last quarter, then write down who signed off on each one. Most centers find the honest answer is headquarters, whoever asked loudest, or nobody documented it at all, not the Center of Excellence itself. That list is the real decision-rights map you have today, regardless of what the charter says.
02 Write a one-page decision-rights matrix and get both sides to sign it
For each of the eight checklist questions above, name the specific role that decides, not the department. Get sign-off from both the GCC head and the headquarters sponsor on the same page, so the next disagreement has a document to point to instead of a fresh debate.
03 Give the center one budget line it fully controls
Pick a modest, well-bounded pool, a single-vendor pilot budget or a defined evaluation spend, and hand full authority over it to the center with no headquarters sign-off required below a set threshold. A center that has never spent money without asking permission has never actually made a decision.
04 Set a fixed review cadence instead of an annual renegotiation
Put a quarterly checkpoint on the calendar where the charter, the budget line, and the decision log all get reviewed together, instead of letting authority quietly erode until the next full renegotiation. Centers that skip this step tend to find their decision rights have shrunk by the time anyone notices.
Let 10decoders Map Your GCC's AI Decision-Rights Gap
Our team reviews what your AI Center of Excellence is chartered to decide against what it can actually approve, fund, or veto today, then hands you a decision-rights matrix both the center and headquarters can sign off on.