When the Knowledge Moat Disappears: What happens to Outsourcing?
If AI democratizes technical execution, what exactly are you outsourcing?
I was in India last month, where the AI conversation has a different urgency than anywhere else I’ve worked.
Four weeks ago, India hosted its AI summit. Last month, Nasscom downgraded profitability forecasts for the Big 3 outsourcing giants—Infosys, Cognizant, and TCS. Then yesterday, Anthropic published research with hard numbers: AI has reached 74.5% coverage of computer programming tasks, with customer service at 70% and data entry at 67%.
The Old Moat Is Gone
For decades, outsourcing meant: “We have cheaper engineers who execute your defined technical requirements.”
That model assumed a knowledge barrier. Writing code required years of training. Building systems required specialized expertise. The moat was access to talent at lower cost.
The Anthropic research shows exactly how that moat is collapsing. Their “observed exposure” measure tracks actual automated usage in professional settings—not theoretical capability. The gap between what’s possible and what’s deployed is closing fast.
Computer programmers show 74.5% coverage. Customer service representatives—whose main tasks increasingly appear in first-party API traffic—hit 70%. Data entry keyers, whose primary work of reading documents and entering data shows significant automation, reach 67%.
But here’s what makes this existential: the same research finds a 14% drop in hiring of workers aged 22-25 into these high-exposure occupations. Companies aren’t just automating existing work—they’re questioning whether they need to hire for routine technical execution at all.
The Big 3’s problem isn’t lacking AI tools. It’s still selling the old value proposition—cheaper execution—when that’s precisely what AI commoditizes.
The Big 3 are being downgraded because they’re caught between models. They can’t compete on cost with AI for routine work. But they haven’t pivoted to selling what AI can’t commoditize.
The Question of Value
Outsourcing companies are using AI to do the same work cheaper and faster. They measure success by “we can deliver your requirements with 30% fewer engineers” or “we reduced project timelines by 40%.”
This is the optimization trap I wrote about in Thursday’s exercises, now playing out at enterprise scale.
They’re making broken processes faster—still broken, just faster. They’re automating execution of already-defined work, rather than asking what becomes possible when you free that capacity.
The question isn’t “how do we use AI to do existing work more efficiently?” It’s “what high-value problems can we now tackle that were previously too expensive?” I started exploring this in one of my first posts (The Vicissitude of Value).
The New Moat
If not cheaper or faster, then what is the new value proposition?
It’s two things:
1. Human judgment at scale
Understanding what problems are worth solving. Knowing where AI can handle tasks unsupervised versus where human oversight matters. Developing fluency to distinguish transformation from theater.
This is the judgment gap I’ve been writing about—at organizational scale.
2. Capacity for higher-value problems
AI doesn’t reduce the need for engineering talent. It changes what that talent should work on.
This requires fundamental reframing: from “we execute your requirements efficiently” to “we identify which problems AI should solve versus which need human expertise—and have freed-up capacity to tackle problems you couldn’t previously afford.”
Those problems that were “too expensive” because engineering capacity was consumed by routine work? Suddenly addressable. That New Initiatives R&D team you could never afford? Create it. Get fresh blood and radical ideas in—invite them to the table.
What This Means for Everyone
This isn’t just an India problem. Western companies hiring outsourcing firms are asking the wrong question too.
The Anthropic research reveals something crucial: AI isn’t eliminating jobs through unemployment (yet)—it’s eliminating them through non-hiring. Overall unemployment rates for high-exposure workers remain flat. But companies simply aren’t filling positions AI can now handle.
If you’re outsourcing routine technical execution, you’re paying for something AI increasingly does better and cheaper. The value isn’t in execution—it’s in judgment about what’s worth building and capacity to tackle strategic problems that create genuine competitive advantage.
The companies that figure this out—whether in India, the US, or anywhere else—won’t be outsourcing firms. They’ll be strategic partners who’ve developed fluency to know where AI creates value versus overhead.
That’s not a technology shift. That’s a business model transformation.
And like all transformation, it requires confronting an uncomfortable truth: the thing that made you successful is exactly what AI commoditizes. The question is whether you’ll reimagine your value proposition or optimize your way to irrelevance.
References: Anthropic Research: “Labor market impacts of AI: A new measure and early evidence” (March 5, 2026) Anthropic Paper here