The AI Shift: What It Means for India's Outsourcing Empire
Exploring the capabilities of agentic AI and the effect this will have on India's current outsourcing empire
Just thought I would quickly introduce myself as a new member of the 2084 blog. Hi, I am Shailen Desai, an eligible Chartered Accountant CA(SA) from South Africa with a passion for automation within the finance industry. I previously worked in the big 4 environment and currently exploring internships in broader finance and IT in Mumbai India
Imagine this: an industry that earns $180 billion annually, contributes a massive 9.3% to India’s GDP, and keeps global giants like Google, Microsoft, and JPMorgan running smoothly. That’s India’s outsourcing empire in a nutshell—a decades-long success story built on IT services, data analytics, and back-office operations.
But here’s the kicker: AI is shaking things up. Tools like GPT are automating workflows, and agentic AI is now making decisions that once needed human judgment. The big question? Can this outsourcing behemoth keep up, or are we seeing the beginning of a huge industry shift?
India’s Rise as an Outsourcing Giant
India’s journey to becoming the world’s outsourcing powerhouse began in the 1990s when Western companies, especially in the United States and the United Kingdom, started moving services offshore to cut costs. Backed by early investments in IT infrastructure and a growing pool of English-speaking graduates, India quickly emerged as the go-to destination for outsourced work.
What makes India stand out? It’s all about the cost-effective, semi-skilled workforce. The numbers paint a clear picture:
22.15% of workers are classified as low-skilled (Skill Level 1).
A whopping 66.89% fall under the semi-skilled category, making up the majority of the workforce driving India’s outsourcing dominance (Skill Level 2)
Meanwhile, only 2.37% are highly skilled (Skill Level 3), and 8.59% fall into the top tier (Skill Level 3)
However, this reliance on semi-skilled labour now faces an existential threat as AI technologies become more capable of performing these tasks faster, cheaper, and with fewer errors.
Quantifying the ability of AI to replace workers
AI is a game-changer, and India’s outsourcing industry is sitting in the splash zone. With tools like Large Language Models (LLMs), AI agents, and **automation platforms**, the tasks that once needed semi-skilled workers are now being handled by computers that don’t need sleep, benefits, or even coffee breaks.
Now, I’ll admit—at first, I didn’t quite get how powerful agentic AI could be. Then, I came across a fascinating study by The Agent Company, which tested these AI agents in roles we’d consider very human: project management, HR, and even finance.
The Agent Company Framework
Here’s how they did it. They set up a fake company—a simulated workspace with all the tools you’d find in a real office:
- A Linux terminal, code editor, and browser for tech-heavy tasks.
- Tools like GitLab (for coding), RocketChat, a Microsoft Teams clone (for communication), and OwnCloud, a Google Drive/OneDrive clone (for document sharing).
Then, they gave the AI agents real-world tasks, grouped into six main roles:
1. Software Development Engineering (SDE): Writing and fixing code.
2. Project Management (PM): Planning sprints and keeping teams on track.
3. Data Science (DS): Analyzing datasets and creating visuals.
4. Administrative Tasks (Admin): Scheduling meetings and managing workflows.
5. Human Resources (HR): Screening resumes and onboarding employees.
6. Finance: Handling reimbursements and financial forms.
Each task had checkpoints (mini-goals), and agents were scored on how much they completed and how efficiently they did so.
The Agent Company Findings
Here’s where it gets interesting:
AI agents crushed tech-heavy tasks like coding and debugging. Why? Tons of training data is publicly available.
But they flopped on tasks like HR and admin. Why? These roles need a lot of understanding, collaboration, and navigating messy systems—things AI isn’t great at (yet).
The best AI agent at the time of testing, Claude-3.5 Sonnet, managed to fully complete 24% of its tasks and make partial progress on another 34.4%. Not bad, but still not close to human-level performance in most roles.
Find below a breakdown of the best open and closed source models per category.
The Future
Here’s the part that excites (and scares) me. While AI might seem a little underwhelming in some areas, that’s about to change. OpenAI’s Stargate project, with a jaw-dropping $500 billion investment, promises to take AI to the next level. Think exponential improvements in understanding, automation, and decision-making.
What does this mean? It means the tasks AI struggles with now—like admin and HR—might not be struggles for much longer. India’s outsourcing industry is about to face a future where AI could outperform humans in almost every way.
My Opinion
The Agent Company’s findings make one thing clear: AI is rapidly advancing and could soon replace workers in semi-skilled and even highly skilled roles. However, if there’s one thing humanity—and India—has proven time and time again, it’s the ability to adapt.
With upskilling, innovation, and a willingness to embrace change, I’m confident India’s workforce will evolve, creating new opportunities and thriving in an AI-driven world. This isn’t the end of the story—just the beginning of a new chapter.
Figure 2: https://arxiv.org/abs/2412.14161