The Death of the BPO: Why AI Agents Are Rewriting Sourcing Strategy

For the better part of two decades, I have worked on one of the most fundamental questions in corporate strategy: what should we do ourselves, and what should we hand off to someone else? The framework was elegant in its simplicity. Identify your core competencies — the capabilities that give you genuine competitive advantage — and protect them fiercely. Everything else? Outsource it. Find a Wipro, an Accenture, a Cognizant, or a TCS, and let specialists handle what you do not need to own. The Business Process Outsourcing industry was born of this logic, and it grew into a $300+ billion global sector. The big players generated revenue in the tens of billions — Accenture alone at $64.9 billion, TCS at $29.1 billion. It was a sound strategy. And it is now being fundamentally disrupted.

The Old Framework and Its Hidden Assumptions

Traditional sourcing strategy rested on a set of assumptions that seemed unshakeable. First: if a function was non-core, you could still execute it well — you just needed the right outside team to do it. Second: the only way to access specialized execution capacity was to hire it, either internally or through a BPO. Third: scale required headcount. More invoices meant more people processing invoices. More customer queries meant more agents in a call center. Every one of these assumptions is now under assault. 

The BPO model was always imperfect. Even when you outsourced well, you were still handing work to human teams operating across different time zones and cultures. The result was predictable: slower turnaround times, occasional miscommunications, and inconsistent quality. The work got done, but handing over work to another human team never guaranteed perfect outcomes. Conventional software like ERP, CRM or ITSM tried to solve this problem but couldn’t close the gap. Rule-based automation handled well-defined, repetitive, structured tasks. But the moment a process required interpreting unstructured data, understanding context, or exercising judgment — things like resolving a nuanced customer complaint or processing a non-standard invoice — the software fell short. The tasks that most needed improvement were precisely the tasks that software could not reliably handle. That gap is what AI agents are now closing.

What Changes With AI Agents

AI agents are not chatbots with a marketing rebrand. They are sophisticated systems that perceive context, make decisions, take actions independently, and interact with external systems — all to accomplish complex business objectives. They can process unstructured data, connect with multiple systems simultaneously, and carry out actions or provide recommendations in real time. This is not incremental improvement. It is a qualitative shift in what automation can do. Consider the HR function: AI systems like IBM’s Watsonx HR agents already handle time off requests, process payroll, administer benefits including healthcare and retirement plans, and manage complex compliance requirements — achieving a 70% reduction in manual tasks and 30% cost savings through automation. IT support agents are saving more than 4,000 calls per month while delivering 24/7 coverage no human team can match. Sam Altman said it plainly: “We believe that, in 2025, we may see the first AI agents join the workforce and materially change the output of companies.” That moment has arrived.

The Strategic Disruption to Sourcing Strategy

Here is the part that executives and strategists have not fully absorbed yet: AI agents do not just make BPO cheaper. They change the logic of the decision entirely. The old question was: is this function core or non-core? If non-core, outsource it to a human team at a BPO. The new question is: can this function be automated entirely by AI agents? If yes, the BPO is no longer the answer — and in many cases, neither is the traditional insource/outsource binary. This reframing matters enormously. It means that many processes previously considered “outsourceable” are now better described as “automatable.” The operational footprint of a company can shrink not because work is moved offshore, but because it no longer requires a human team at all. AI agents expand operations without a matching rise in staffing. For businesses with volume-based cost models — say, invoice processing that scales with sales — AI becomes especially attractive because it flattens or eliminates that cost curve entirely. The sourcing strategist’s toolkit needs a new column.

Who Gets Disrupted, and Who Gets Empowered

The disruption falls unevenly. Large BPO companies are not oblivious to this shift — KPMG has committed $2 billion in AI investments with Microsoft, PwC $1 billion, Deloitte $1.4 billion, EY $1 billion — but their core business model is still built on human labor. Transitioning from service-based revenue to product-based revenue at that scale is genuinely difficult. The very size that made them dominant now creates structural inertia. Meanwhile, two groups stand to gain significantly. 

The first is the mid-sized enterprise that previously could not afford sophisticated BPO services. AI solutions open “BPO-style capabilities” to organizations that were priced out of the market, democratizing access to advanced operational capabilities. A company with 200 employees can now deploy HR automation, financial analysis, and customer support infrastructure that only a Fortune 500 could previously access. 

The second is the large enterprise willing to rethink its operational footprint from scratch. AI agents can scale into new product lines or extra back-office functions without headcount increases. This frees executive bandwidth to focus on genuinely strategic and creative work — the kind of work that actually requires human judgment and relationship capital.

The New Sourcing Strategy Framework

Given all of this, I would argue that the classic sourcing decision framework needs a third option. It used to be:

1. Insource — if it is a core competency
2. Outsource to BPO — if it is non-core Now it should be:
3. Insource — if it is a true core competency requiring human judgment and strategic ownership
4. Automate with AI agents — if it is a non-core, process-driven function that AI can handle reliably
5. Outsource to BPO — for complex, judgment-intensive non-core functions where AI is not yet mature enough, or where the transition cost outweighs the benefit The middle option — AI-agent automation — will expand rapidly. 

What This Means for Leaders Right Now

If you are a Chief Operations Officer or Chief Strategy Officer, the implications are practical and immediate: 

1. First, audit your sourcing portfolio with fresh eyes. For every function you currently outsource, ask honestly: could an AI agent do this? The answer is more often “yes” than it was even twelve months ago. 
2. Second, do not wait for your BPO provider to lead this transformation for you. Their incentive is to preserve headcount-based revenue. Your incentive is efficiency and competitive advantage. Those are not always aligned. 
3. Third, resist the temptation to treat AI deployment as a purely technical decision. The most important barriers are organizational: redesigning job functions, reskilling people whose roles are automating, and redefining what human-AI collaboration looks like in practice. Technology is the easy part. 
4. Finally, recognize that the companies that move earliest and most deliberately will establish structural cost advantages that are very hard for competitors to close later. AI costs are plummeting — powerful reasoning models have been trained for under $500, sometimes under $50 in cloud credits. The democratization of capability is happening now.

The BPO industry will not disappear. There is still significant value in experienced human judgment for genuinely complex, relationship-intensive, or highly regulated work. But the center of gravity is shifting. The playbook that has guided sourcing strategy decisions for twenty years is overdue for a rewrite. The question is no longer simply “should we make or buy?” The question is now “should we make, buy, or automate?” And for most processes, the answer is increasingly: automate.