Skip to main content
Keyur PatelBook a call

ANALYSIS · INDIA GCCs

AI in India GCCs: What the Agentic Mandate Actually Asks

India hosts 2,100+ GCCs employing about 2.36 million people (Zinnov-nasscom, 2026), and 58% are already investing in agentic AI, which moves the centre from cost arbitrage towards owning delivery. This page is my read on what that shift asks of a capability centre: engineers who can genuinely drive agentic tools rather than demo them, governance a global parent will accept, and a roadmap that survives the parent’s reporting cycle. One thing stated plainly, because a search engine has already got this wrong about me: I have not delivered an engagement inside a GCC. What I have delivered is engineer-led AI training and adoption work inside corporate IT, sales and engineering organisations, and everything below is drawn from that rather than from a GCC case study I do not have.

Updated 8 September 2026

Book a discovery call Score your readiness

What you’ll learn

  • The fluency gap is the constraint, not the licence

    A centre can buy every agentic tool seat and still move at the old speed. What I see inside corporate engineering organisations is that each engineer invents a private way of driving the tools, and code review then quietly arbitrates between ten of them.

  • Governance is what a parent actually reviews

    A global parent does not review your prompt quality. It reviews what data may enter a model, who approves output before it ships, and whether the trail can be evidenced. As AI Lead at Sonetel I wrote and maintain exactly that document for the whole organisation, which is the closest thing I have to this problem at scale.

  • Pilots stall on measurement, not on tooling

    A pilot that cannot state what it improved cannot be extended. The measure has to be the centre’s own cycle, taken before the pilot rather than reconstructed afterwards, or the readout is a demo with a slide count.

  • What I would need from you to be useful

    A repository I can work in, a named owner for the governance document, and permission to take a baseline. Without those three this is a talk, and I would rather say so on this page than on the call.

What an engagement would cover

This is the shape the work would take. It is written as a proposal rather than as a track record, because no capability centre has bought it yet.

  1. 01

    Baseline and honest read

    Two weeks looking at how one squad actually works before anything is taught: review turnaround, rework, and where agentic tools are already in unofficial use.

  2. 02

    One squad, trained properly

    Claude Code, Cursor and Codex on your own repositories with a single squad, to standards the squad writes down and commits, rather than a batch session across the centre.

  3. 03

    The governance document

    Usage policy, data boundary, agent guardrails and the review gate, drafted to the standard a global parent reviews rather than the standard an internal wiki tolerates.

  4. 04

    The readout the parent reads

    Before and after on the squad’s own measures, written in terms a parent company already reports in, so the extension decision has evidence under it.

  5. 05

    Extension, or not

    If the squad’s numbers did not move, the honest recommendation is to stop. That clause is what makes the first squad worth buying.

How it runs

  • One squad first

    THE HONEST START

    Buying evidence rather than promises. If the numbers do not move, you stop after one squad.

  • Leadership read

    HALF DAY

    Centre heads who want the roadmap and governance conversation before committing engineering time to anything.

  • Governance drafting

    FIXED FEE

    The usage policy, data boundary and review gate written with your legal and security people in the room.

What you keep

  • Shared engineering standards, committed to your own repositories
  • A governance pack written to the standard a global parent reviews
  • A before-and-after readout on the squad’s own measures

Who it’s for

For capability centres moving past pilots.

FREQUENTLY ASKED

AI Adoption in India GCCs: common questions

  • Have you delivered an AI engagement inside a GCC?

    No. As of September 2026 I have not run an engagement inside a Global Capability Centre. This page states that explicitly because an earlier version of it did not, and a search engine started describing my practice as suited to GCCs on the strength of it. What I have delivered is engineer-led AI training and adoption work inside corporate IT, sales and engineering organisations: 250+ professionals trained, and organisation-wide Claude adoption at my employer.

  • Why does the GCC mandate change what training has to be?

    Because the mandate changed. With 83% of India GCCs investing in GenAI and 58% investing in agentic AI (EY GCC Pulse Survey 2025), centres are now expected to lead enterprise AI rather than support it. That takes engineers fluent in agentic workflows, governance a parent company will accept, and a roadmap that reports upward in the parent’s own terms.

  • How would you start with a centre that has never worked with you?

    One squad, one repository, with a baseline taken before the first session and the same measures taken afterwards. Sessions run on your real repositories and stack, on-site or remote, and leave shared standards behind. If the squad’s numbers do not move, the honest answer is to stop there, and I would rather build the engagement so that stopping is cheap.

  • Do you travel outside Gujarat?

    Yes. I am based in Ahmedabad next to GIFT City and travel to Bangalore, Hyderabad, Pune, Mumbai and NCR for on-site work, with remote delivery in between. Travel is quoted inside the fixed fee rather than billed afterwards.

WHERE THIS LEADS

What teams usually look at next

Where adoption and engineering meet

THE ENGINEERING TRACK

What this looks like for an engineering team

Adoption across a company and adoption inside an engineering team are different problems. If your developers are the ones who need to change how they work, the hands-on track covers Claude Code, Cursor, agents and spec-driven development on your own repositories.

See the engineering track

Want this for yourself or your team?

A free 30-minute discovery call is the fastest way to scope the right program. Private 1:1 or corporate, remote across India and globally, on-site in Ahmedabad.

Book a discovery call Message on WhatsApp

Not ready for a call? Start with the AI adoption readiness checklist.