DEPT’s Andrew Dimitriou Bets India Outruns Automated Back Offices

Andrew Dimitriou told a June 8, 2026 interview that India should export new AI-enabled services, and that the old back-office model will keep existing only as it is automated. The same week, the Global Chief Client and Growth Officer at DEPT said the agency had already helped some clients cut content production costs by as much as 75%.

That pairing is the wager. India now runs a GCC engine of 2.36 million people. The work those centres still sell is the work his cost cut describes.

A 75 Percent Cost Cut Is Also a Staffing Problem

Dimitriou joined DEPT on September 16, 2024 as a partner, moving from seven years as EMEA chief executive at VMLY&R, where he led a network of 4,500. He reports to global chief executive Dimi Albers and sits in Europe. The Dutch-founded shop, backed by the Carlyle Group since January 2020, said at his appointment that half of DEPT’s €500 million-plus revenue was already AI-enabled, after revenue growth of over 200% in three years.

In the June interview he split AI into two piles. Repetitive, rules-based tasks should be automated. Empathy, culture and creative judgment should not. Brands that pour the same models onto every brief, he said, will look, sound and behave like everyone else.

We’ve helped some clients reduce content production costs by as much as 75%. At the same time, those savings are being reinvested into new channels, new experiences and new growth opportunities. AI lowers costs, but it also raises expectations.

Andrew Dimitriou, Global Chief Client and Growth Officer, DEPT

He called the biggest mistake in the market a fixation on efficiency, because the same tools will be available to everyone. “The biggest opportunity in AI isn’t productivity. It’s reinventing growth.” Companies that stop at a cheaper version of last year’s process, he said, will be picked off by AI-native rivals.

That is also why so many firms never leave the lab. Dimitriou said they treat AI as a technology project, then bolt it onto a decades-old operating model. New companies skip the legacy incentives. Established ones often cannot. The stall shows up as AI pilots that stall after the demo, which he traced to human behaviour rather than model quality.

India’s GCC Engine Hits $98.4 Billion

The Nasscom-Zinnov landscape study for FY2026 puts India at 2,117 GCCs and $98.4 billion in revenue, spread across 3,728 units. Centre count is up 32% since FY21, when revenue was $61.4 billion and the workforce was 1.7 million. About 506 Forbes Global 2000 companies now run a centre in the country.

THE FY2026 GCC SNAPSHOT

• The headcount: 2.36 million professionals, up from 1.7 million in FY21.
• The AI layer: more than 1,200 centres embed AI and machine learning, with 250 dedicated centres of excellence and 250,000 AI professionals.
• The mandate shift: 96% of centres opened after FY21 launched with a product or portfolio brief, and nearly 50% of all GCCs now sit at high maturity.
• The new-build mix: nearly half of GCCs set up since FY21 were built with AI at the core.

Nasscom president Rajesh Nambiar said the reset is already running, with AI as the catalyst, and that centres are taking ownership of global products, platforms and business outcomes. The public GCC conversation often treats that sentence as finished business, as if execution work had already left the building.

Hiring data from after Dimitriou’s interview is less tidy. Naukri’s JobSpeak index for August 2026 showed AI and machine learning postings up 31% year on year and GCC hiring up 10%, with overall white-collar hiring up 14%. Hyderabad led GCC hiring growth at 28%, then Chennai at 21%, Pune at 15% and Bengaluru at 13%. The jobs arriving are not the jobs that built the back office.

Jobs platform foundit told a trade survey that AI-linked roles in India’s media and communication GCCs rose from 21% of jobs in 2024 to 44% in 2025. Campaign reporting, media trafficking and low-complexity execution support slowed as those workflows moved into software. Holding-company networks kept adding India seats anyway. The mix inside those seats is what changed.

Why Agencies Are Being Told to Stop Selling People

Dimitriou’s prediction, the one he said many agency leaders would reject, was blunt. “The future of agencies isn’t selling people. It’s selling outcomes.” Clients, he said, will buy capabilities and business results rather than headcount. Firms that stay on time-and-materials will be treated as a commodity.

He has already watched the commercial model move toward output-based and outcome-based fees. If production that once filled buildings can fall by as much as 75%, hours stop being the product. The remaining scarce inputs are the ones he listed first: strategic judgment, cultural relevance, and the ability to turn thousands of machine-made interactions into one brand.

New work is appearing in the gap. “Six months ago, brands weren’t building shoppable experiences inside ChatGPT environments. Today, they are.” The date on that remark is June 8, 2026. The window he was pointing at opened in late 2025. Agencies still organised around traffic, versioning and reporting will not automatically inherit those briefs.

He also warned that volume is the wrong scoreboard. Brands are making more content than they can use. The fight, he said, is usefulness: the right piece, in the right context, to the right audience, at the right moment. Distribution intelligence matters. It does not replace the judgment that decides what is worth sending.

Cheap Content Makes Brand Sameness Easy

If anyone can generate copy, the scarce resource is relevance. Dimitriou’s short answer to the abundance question was a single word, judgment, then a longer one: competitive advantage. Orchestration beats output.

Three days before that interview, on June 5, 2026, DEPT put an AI content engine, DEPT Studios, into India on Adobe’s stack. Dimitriou said then that the best creative teams want to do more than production capacity allows, and that “DEPT Studios changes that equation and nowhere is that truer than India.” Alison Webster, vice president of Adobe’s strategic partnership program, said brands expect content demand to grow 5x by 2027, because customers respond to highly personalised work.

India is the test kitchen for that claim. Dimitriou called the country a microcosm of the world, with multiple languages, regions, cultures and digital systems in one market. Hyper-personalised content at that scale was, until this generation of models, a fantasy briefing slide. Done badly, the same stack will emit fluent mush in twelve languages.

He said authenticity now depends less on whether the model can write and more on the safeguards around what consumers actually see. The technology, in his view, has matured. Process has not. Brands that skip the process will not look innovative. They will look interchangeable.

DEPT Already Runs a Large India Shop

The growth officer making the hub argument is not talking about a market DEPT does not staff. On a July 2025 visit with Vishnu Mohan, partner and chief executive for Asia Pacific, Dimitriou called India one of the firm’s largest operations by headcount. He was meeting clients, prospects and the local team, and said every conversation drifted toward what technology was doing to marketing. He also said DEPT’s top 25 clients were growing at 10% under a new global key-account program, with partnership bets on Adobe, Salesforce, Shopify, Google and OpenAI.

DEPT’S PATH INTO INDIA

1. February 1, 2023: Tekno Point’s 500 Adobe specialists join DEPT. The Mumbai firm, founded in 2000 by Himanshu Mody and Yash Mody, brought Tata Capital, Bajaj, Asian Paints, Rosewood Hotels and AIA on Adobe work. Albers called India a world-class talent pool and a domestic market with “enormous potential.”
2. March 3, 2025: DEPT opens a Bengaluru hub under Rishi Bhargava, head of solution consulting. The India team is over 700, with clients including HSBC, Tata Capital and Bajaj Allianz. Himanshu Mody, then head of India, said the city move answered demand from brands that wanted to disrupt their categories.
3. June 5, 2026: DEPT Studios launches in India. The content practice had grown 15% over the prior year, with plans to add 50 roles through 2026.
4. June 8, 2026: Dimitriou argues that India can lead the next wave of global marketing innovation, and that the back-office model will be automated.

DEPT lists more than 4,000 people across 30-plus offices on five continents. The India bench is large enough that a speech about leaving execution work behind is also a speech about what those seats will do next. Mohan has described the local design practice as international craft fused with domestic talent, with mentoring from the Netherlands. That is a hub story. It is also still a delivery story until the billing model changes.

Model-Agnostic Layers Beat a Single Lab

Dimitriou’s answer to the foundation-model bottleneck was to refuse a single vendor. The strongest strategies, he said, will be model-agnostic. Different models for different jobs. Companies should keep their data, workflows and proprietary tools, then sit an intelligence layer on top that can call more than one lab.

That matters in India because so much GCC work still rides on platforms owned elsewhere. A centre that only operates someone else’s model is still a back office, even if the slide says AI. A centre that owns the workflow, the data and the orchestration is selling something that does not disappear when a lab ships a cheaper API.

THE SPLIT DIMITRIOU IS DRAWING

• Efficiency shops: they automate reporting, trafficking, versioning and other rules-based tasks, then stop. Everyone else will buy the same tools.
• Growth shops: they spend the savings on new channels, shoppable agent environments and services that did not exist a year earlier.
• Model renters: they depend on one lab and do not keep the data or the workflow. The strategic risk sits with the vendor.
• Layer builders: they combine models, keep the intelligence layer, and bill for outcomes rather than seats.

He put the same split on talent. AI replaces tasks first. Jobs follow if people do not move with the tasks. Retraining is not a slogan in that frame. It is the only way a 2.36 million-person machine does not throw off a generation of process specialists with nothing left to process.

Hiring Still Chases the Old Execution Product

When Dimitriou was asked how GCC leaders should rebuild pipelines designed for scale and process efficiency, he said they have to move from outputs to outcomes. Historically the brief was to deliver work cheaply. In an AI shop the brief is to solve a business problem. The winning mix, he said, is AI efficiency paired with human intelligence.

Workforce firm Quess Corp’s “GCC 2.0: The Rise of Digital Business Functions” report, out September 4, 2026, put numbers on what that looks like when the old product is withdrawn. Demand for routine roles fell across all eight major non-tech job families in 2025. The sharpest drop was in procurement.

ROUTINE GCC DEMAND, 2025 VS 2024

Non-tech GCC demand is not disappearing. Quess counted about 339,200 such roles in 2025, up 49.2% year on year, and projected about 500,000 by 2028. Sales operations, finance and business operations still make up over 60% of that book. The people being hired are mid-career: professionals with 7 to 12 years of experience were nearly 60% of non-tech demand. Nearly two in five business-operations postings, and one in two analytics postings, now ask for digital platforms, analytics or AI tools as standard.

That is the quiet version of Dimitriou’s bet. India can still add seats. The seats that used to be sold as cheap, repeatable output are the ones going away. Nasscom’s own outlook for the next five years is a move from talent scale to expertise density, and from cost arbitrage to AI-anchored relevance. Dimitriou’s version is shorter. Capability is not the question. Willingness to leave the execution model is.

He would bet, he said, that India becomes something larger than the world’s back office, because that model will be automated and the export prize sits in new AI-enabled services. The 75% content cut is already on the invoice. The procurement drop is already in the hiring file. The hub exists only if the next brief is a business problem, not a cheaper version of the last one.