AI, GCCs, and the Reshaping of IT Outsourcing

AI, GCCs, and the Reshaping of IT Outsourcing

IT outsourcing is going through its sharpest structural change since offshore delivery became standard. Large services firms have cut tens of thousands of roles globally this year, with a substantial share landing in Indian delivery operations, while hiring projections for the coming half year point downward again. At the same time, global capability centres, the captive units multinationals run in the same cities, are hiring at record levels and expanding well beyond technology into finance, analytics, and research roles. Both things are happening in the same labour market, often within walking distance, and together they describe a shift in who does the work rather than a simple contraction.

Third-Party Provider or Captive Centre

The trade-offs have shifted, but they have not disappeared.

What Is Actually Happening to IT Outsourcing

The headline reading, that AI is replacing outsourced workers, is too simple. What is being compressed is a specific layer: high-volume execution work priced by the hour or the seat, where the task is well defined and verifiable. Ticket triage, routine test execution, first-line support, and standard code maintenance sit squarely in that layer, and they are the easiest to automate meaningfully.

Above that layer, demand looks healthier. Work that requires context about a client's business, judgement under ambiguity, regulatory knowledge, or accountability for an outcome is not automating away. Industry analysts have argued that the sector holds its ground by moving up the value chain rather than by defending volume, which is consistent with the hiring pattern: fewer delivery roles, more specialised ones.

Why AI Hits the Execution Tier of IT Outsourcing First

Because that is where the economics are cleanest. The work is repetitive enough to automate, measurable enough to prove savings, and priced in units that make the saving visible immediately. It is also the layer that was most exposed anyway, since its only defence was cost arbitrage, and arbitrage does not survive a technology that reduces the hours rather than the hourly rate. Anyone whose contribution is bounded by a ticket queue is in a harder position than someone who owns an outcome.

Geography is also less decisive than it used to be. Once a task is automated, the cost of performing it stops depending on where the person doing it lives, which removes the original logic for moving it offshore at all. What still makes a delivery location valuable is talent depth, time zone coverage, and regulatory fit, and those are very different criteria from hourly cost.

Why Captive Centres Are Absorbing the Shift

Multinationals are internalising delivery for reasons that predate this year. Proximity to product teams shortens feedback loops, retained intellectual property matters more as software becomes the product, and AI work in particular depends on access to proprietary data that companies are reluctant to place with a vendor. Hiring data shows the fastest growth in analytics, data science, and AI roles, which is exactly the profile a company wants to own rather than rent.

That said, captives are not free. A capability centre converts variable cost into fixed cost, imports management overhead across time zones, and becomes very difficult to unwind if strategy changes. The organisations doing this well are selective, keeping a captive for the capabilities that define them and continuing to contract for everything else.

The hybrid structure is where most large buyers are landing. A captive holds the capabilities that define the business, a small number of strategic providers carry work that benefits from scale and specialist tooling, and commodity delivery goes to suppliers that are easy to replace. That arrangement takes more management effort than a single relationship and it is far easier to adjust when the technology moves again.

Six Questions for Buyers Renegotiating Now

Most current contracts were written for a delivery model that is disappearing.

1. What are we actually buying? Separate outcome-based work from staffed capacity before discussing rates.
2. Who captures the automation benefit? If the provider automates a workflow, agree the split rather than leaving it to the next cycle.
3. Which capabilities must we retain? Decide deliberately which knowledge should never sit only with a vendor.
4. How is quality measured now? Headcount and handle time stop describing value once work is automated.
5. What happens to our data? AI-assisted delivery changes where information flows and which models see it.
6. What does exit look like? Transition assistance written for a staffed model may not fit an automated one.

Pros and Cons of Moving Work In-House

Bringing delivery inside solves real problems and creates new ones.

• Pro: control of data and intellectual property. Proprietary information stays inside the company boundary.
• Pro: shorter feedback loops. Teams sit closer to product decisions and customers rather than behind a contract.
• Pro: capability compounding. Knowledge accumulates with people you employ instead of leaving with a supplier.
• Con: fixed cost. A centre must be staffed through downturns as well as growth phases.
• Con: management load. Running operations across time zones is a skill many buyers underestimate.
• Con: hard to reverse. Closing a captive is slower, costlier, and more visible than ending a contract.

Real Scenarios Worth Thinking Through

These scenarios are illustrative, showing how the shift in IT outsourcing plays out in practice rather than presented as verified case studies.

A bank renews a support contract at a lower seat count after automation lands, then discovers that the residual work is almost entirely complex exceptions. Average handling time rises, the provider staffs it with more senior people, and the effective unit cost is higher than before even though the invoice is smaller.

A retailer builds a capability centre for data and analytics and keeps its infrastructure and support contracts with providers. Two years later the analytics team is a genuine advantage and the support relationship is unremarkable, which is exactly the intended outcome of choosing what to own.

A manufacturer lets a provider use AI tooling across delivery without changing the contract. Productivity gains accrue entirely to the provider, and the buyer only notices during benchmarking, when its rates look unchanged while the market has moved.

“After testing multiple AI models on coding, research, and business prompts, combined outputs produced more reliable results than any single model.” Internal multi-model evaluation, Talkory research team.

What Providers Are Doing About It

The serious providers are repricing rather than defending volume. That means moving toward outcome and transaction pricing, building domain platforms they own, and pushing consultants up the stack into advisory and orchestration work. Some are also selling AI governance and verification services, which is a reasonable position for a firm that already sits inside a client's processes.

Buyers should read those moves carefully. A provider that shifts to outcome pricing and invests in its own tooling is adapting. A provider that offers the same staffed model at a slightly lower rate is managing decline, and its ability to invest in the capabilities you will need next year is limited. The commercial questions behind that shift are covered in our piece on BPO AI agents and seat pricing.

Why Talkory Wins

When delivery becomes AI-assisted, the buyer's assurance problem changes. You are no longer checking whether enough people were assigned. You are checking whether the output is correct. Talkory lets a vendor management team take a sample of delivered work, run the same task across GPT, Claude, Gemini, Grok, Perplexity Sonar, and Kimi K3, and compare. Convergence suggests the delivered answer is sound. Divergence identifies the question types where automation is being used beyond its competence, which is precisely the evidence that makes a commercial conversation concrete rather than anecdotal.

Final Verdict

IT outsourcing is not collapsing. It is losing the layer that existed because labour was cheaper somewhere else, while captive centres absorb the work companies now consider strategic. For buyers the useful question is not offshore or onshore, but which capabilities must live inside the company and which can be bought. Rewrite contracts around outcomes, agree who captures automation gains, measure delivered quality rather than staffed capacity, and decide deliberately what you keep. The organisations that answer that clearly will spend less and rely on their suppliers more effectively than those still negotiating rate cards.