India's Rise in AI Outsourcing​

Contents

With every can of worms comes out a caterpillar. As the tech industry digested the news of Amazon pulling back its AI-driven 'Just Walk Out' shopping format, that is, stores without a conventional cashier, conjectures about a new 'Mechanical Turk' began to emerge. Even as the e-commerce giant refuted these claims, questions popped about who (and not what) exactly powers the AI offerings that are being doled out at such a brisk pace from BigTech kitchens.

Ironical as it is: Using human "AI" talent to steer AI algorithms and models that kill human inefficiencies could be the new reality. And India could be a big part of this. But would it be low-brow work? The new debate on 'is AI actually AI' spurs an interesting question. Is India really turning into an AI-sourcing hub? Is it because of those numerous decades of outsourcing lineage? And most importantly, is it just 'grind work' and 'grunt work' or a new stronghold of expertise that is way beyond old-school outsourcing?

Why India? Why AI?

Let's see it from the prism of an IT services firm that has been serving global clients for many years now. And we find out why things are moving upstream this time instead of the long-tail work that used to come to Indian shores earlier. While traditional outsourcing typically involves delegating specific tasks or processes to external vendors or service providers, the involvement of Indian talent in AI products goes beyond mere execution, enlightens Sundar Ramaswamy, Head of Analytics CoE at Happiest Minds Technologies. "Indian talent is actively engaged in ideation, innovation, and the creation of AI products and solutions. Rather than just executing predefined tasks, Indian professionals are contributing to the design, development, and advancement of AI technologies, often, leading to the creation of groundbreaking products and services. Additionally, many Indian startups and companies are building their own AI products and platforms, further distinguishing their involvement from traditional outsourcing models."

Vasupradha Srinivasan, Principal Analyst at Forrester, delineates that these are emerging needs from the accelerated adoption of AI across the globe. "This is different from traditional outsourcing both in terms of the nature of workloads and the quality of talent that is needed to manage these newer work types", she argues. If we look at AI outsourcing from US and Europe, then all large Indian IT companies are now working on AI solutions for their clients.

Global tech majors such as Microsoft and Google have also been bullish about India, as Devroop Dhar, Founder of Primus Partners, points out. There's more to this wave. Incidentally, as observed in a report by NASSCOM-BCG (National Association of Software and Service Companies) report, the demand for AI services alone could be worth 17 billion US-dollars by 2027 in India. That year could also see a global AI market worth 320 to 380 billion dollars by 2027 with the software and services segment holding about 88% of the market. The February 2024 NASSCOM-BCG report, which looks at AI exemplars amongst Indian technology service providers in, shows that 30 percent of players are offering Generative AI advisory and custom model fine-tuning services for domain-specific solutions, while 25 percent of the companies have been able to build a sizable AI (over 20%) and Generative AI (over 5%) client portfolio.

Just two examples: Fraud detection for business process outsourcing, where AI algorithms observe every decision point of the value chain. And a real time speech-to-speech translator to support Japanese and US based teams of an insurance company, which helps reduce the need for manual translation.

The industry in India has already witnessed more than 15 percent growth in AI/ML jobs in the last 12 months. Positions like AI engineers are even growing at over 70 percent year over year.

Is there really a Turk?

Now that we know why India makes sense from the lens of expertise and market potential, let's reckon what kind of work is being outsourced. Is this work back-end work or work needed before complete AI automation is achieved, that most big AI products promise? That is some of the work, yes, avers Srinivasan: "But there are more AI-aligned services too. For example, data cleansing and preparation is back-office work to get enterprise data to be AI-ready, services like building AI and ML models for analytics, building conversational interfaces based on Generative AI, improving the performance of conversational interfaces like chatbots, making enterprise knowledge more accurate and user-friendly with Retrieval Augmented Generation (RAG) models et cetera; these are very different from traditional back-office workloads and are services on their own", she contends.

Ramaswamy echoes that view. "The shift from cost arbitrage to value enhancement has been a constant theme in the knowledge services space and more so in AI/Analytics." In today's post-COVID India, Global Capability Centres (GCC, also known as Global Captive Centres) are no longer mere back-office support for multi-national companies, but rather key to building global solutions and generating additional revenue streams, reasons Ravinder Pal Singh, deep-tech investor and co-founder of the School of Entrepreneurship at Rishihood University. Krishna Vij, head of IT staffing at TeamLease Digital, notes how unlike the routine, repetitive tasks typical of traditional outsourcing, current engagements focus on high-value, strategic work such as research and development.

The Twist This Time is Talent

Thus, unlike previous outsourcing 'vantage points' that the industry in India has enjoyed, this time round, the edge is not in numbers (or as some used to tag it, in 'body shopping') but in real AI prowess. Companies are seeing India as a large consumption market as well as a large talent tool, underlines Dhar: "About 16 percent of today's AI talent is in India. India is among the top 3 talent pools for AI, and therefore no development is possible without involvement of the Indian IT industry." Indeed, when we look at top 5 countries for AI talent (as per Linked Insights-BCG analysis), India emerges at the peak – whether it is Large Tech (75%), Mid Tech (42%), or business process outsourcing (69%).

It is not a co-incidence that AI leadership hiring saw a 15 percent jump while AI engineers' hiring rose by about 70 percent in the last year. India has been very serious about investing in AI-skills, concurs Srinivasan: "Many service providers have already begun enabling their workforce with trainings, learning programs, and certification for AI skills to work well with AI-powered solutions, to manage prompts and interactions with AI-powered tools, and data management to enable efficient use of data." Indian talent and capability centres play a significant role in driving AI innovation in the US and Europe, as noted by Vij: "Their contributions encompass essential back-end tasks like data annotation and pre-automation development, shaping the future of AI globally. India emerges as a powerhouse for leading tech companies, driving crucial AI projects and patent filings. Additionally, this work is also laying the base for complete AI automation by developing advanced AI algorithms, improving data quality, and integrating AI with other technologies. This preparatory work is crucial for achieving seamless AI automation, and Indian tech professionals are playing a vital role in driving this progress for various organizations across markets."

Dhar augurs that India is slowly becoming the AI development hub, with about 16 percent of the total AI development workforce being in India: "It is estimated that about 1 in 4 Git Hub project on AI comes from India. With the availability of significant talent, a large part of AI development work is getting outsourced to India." In 2022, India was at the forefront of GitHub, accounting for a significant 24.2 percent of AI projects. The European Union and the United Kingdom, together contributed 17.3 percent, while the United States was responsible for 14 percent.

What's in The Plot Next?

Watch out as India is emerging as a hub for data annotation services, which is a crucial component of any AI solution with large language models (LLMs), Dhar foretells. "It is an important step in building datasets to train AI models or to keep the AI model safe. India is expected to be a leader, and would employ about one million people by the end of the decade for this." It is also reassuring, that about 30 percent of service players have a comprehensive security policy for AI; building the right guardrails, fine-tuning models and transparency.

But also watch out for the flip side. As Singh sees it, this trend is however a boon as much as it is a curse, as the leading players in the GCC market may have to cannibalize their existing revenue streams, sacrificing short-term profits for long-term capital investments and R&D, redirecting the necessary talent and resources to enable innovation, while India is still lagging behind the global leaders. While India has the advantage of strong population growth, India faces the challenge of raising the large investments required to build corresponding infrastructure and the sheer volume of computing requirements.

"For example, one of Microsoft's largest R&D Centres outside the USA is the India Development Center located in Hyderabad, where basic and applied research is conducted in areas such as AI, cloud computing, data analytics, NLP, and Quantum computing", says Singh, "The physical infrastructure and computing power originates in the US." Singh also unravels that the country still needs to raise significant investment funds towards computing and data centres which can then be routed towards indigenous development. Until that happens, India will continue to play the role of an innovation supporter rather than an innovation creator.

India is trying to to bring its own recipe to the AI party too. Ramaswamy points to startups like Zoho, Freshworks, and CureFit. Ankush Sabharwal is the Founder and CEO of CoRover, the company behind BharatGPT, a Generative AI product from-and-for India. "Most large language models use Western language and conversational data", reminds us Sabharwal, "We know how Indians ask questions and what they ask. So the LLM needs a different approach."

Turns out that India could be making its way to the front row after being in the shadows of outsourcing projects for long.

(ds)