The Next Wave of Offshoring Isn't AI Replacing Offshore. It's a Fight Over What's Left.

Every offshoring decision of the last three decades was chasing the same number: cost per contact. Nearshore, then offshore, then further offshore: each wave was the same trade in a cheaper location. So, when AI arrived to manage customer contact without paying anyone, anywhere, the obvious next question wrote itself: is offshore-to-AI simply the next, cheapest wave in the same sequence?

The data says no, but not for a comforting reason. It says the sequence itself has changed shape.

What the claim is not

This is not an argument that offshore business process outsourcing is safe, nor that AI will hollow it out wholesale. Both of those are too clean, and the evidence doesn't support either. The precise claim is this: AI is taking over a specific, high-volume slice of contact-centre work (the routine, well-defined, scriptable interactions) regardless of where that work currently sits geographically. What's left after that slice is removed, is smaller, harder, and worth more per hour. The real contest isn't "AI versus offshore." It's over who ends up owning that smaller, harder, more valuable remainder, and whether it stays in the location that built the operational expertise to manage it, or gets pulled back onshore.

Why this isn't hypothetical

The Philippines is the clearest place to watch this play out, because nowhere else has as much riding on the answer. Its business process outsourcing sector employs somewhere between 1.6 and 1.8 million people and generates on the order of $38 billion a year, a sector that accounts for only around 3 to 4 percent of total employment but roughly 7 to 8 percent of GDP, a size the IMF has flagged as "macro-critical" precisely because a shock there doesn't stay contained to the sector.

Estimates of exposure vary by study and method, which is itself worth being honest about rather than picking the most dramatic figure: IMF analysis has put the share of Philippine jobs at high risk of AI displacement at roughly 14 percent overall, with other IMF-linked estimates for BPO specifically running as high as a third, and the Philippine Department of Trade and Industry citing a figure closer to half of BPO roles. The range is wide because "displacement" is being measured differently across these studies; some are measuring task exposure; others are measuring job-level substitutability. What's consistent across all of them is the direction: this is the most exposed major employment sector in the country.

What's also consistent, and less widely reported, is that the sector isn't simply shrinking under this pressure. A "BPO 2.0" model is visibly emerging across Philippine operations: AI now absorbs the high-volume, low-complexity interactions (password resets, order tracking, balance enquiries), while agents are repositioned onto the complex, emotionally sensitive and relationship-dependent conversations that remain genuinely hard to automate. Early benchmarks from AI-augmented Philippine operations suggest agents in this model now spend roughly 85 percent of their time on higher-value work rather than repetitive tasks, and that AI-certified agent roles command 30 to 50 percent more in compensation than the roles they replaced. The national outsourcing body, IBPAP, has committed to spending at least $25 million a year on workforce reskilling, alongside a government-backed programme (Project UNLAD) worth roughly ₱740 million and a dedicated AI training academy opened in Cebu in early 2026. This is not a sector waiting to be automated away. It's a sector actively trying to become the thing that operates the automation.

There is a second front adding real urgency to that effort: proposed U.S. legislation, including the "Keep Call Centers in America Act" and the "HIRE Act", aimed specifically at making offshore customer-service delivery less commercially attractive for American companies, regardless of what AI does. Philippine BPO leadership is now responding to AI exposure and onshoring-driven political pressure simultaneously, on a compressed timeline, which is why three separate industry-body restructurings happened within three weeks of each other in June 2026.

The pivot: this was never really about geography

Read the IMF's own more granular finding and the "AI eats offshore" story gets harder to sustain. Of the BPO roles it classifies as highly exposed to AI, a majority, around 61 percent, are also classified as highly complementary, meaning the more likely outcome for those specific roles is AI augmenting the person, not replacing them outright. That matches what the BPO 2.0 data shows on the ground: the work is being redistributed, not eliminated, and the people best placed to do the redistributed work are the ones who already understand the operational complexity, which, for now, is disproportionately the existing offshore workforce, not a fresh onshore hire.

That is the pivot senior leaders should be tracking. The AI layer needs an orchestration layer around it: someone who owns escalation, exception-handling, quality, and the judgement calls a language model still gets wrong often enough to matter. That orchestration role is up for grabs. It can be captured by the offshore provider repositioning itself up the value chain and pricing accordingly. Or it can be pulled back onshore, or in-house, by the client, leaving the offshore relationship holding only whatever thin, commoditised, price-competed volume is left once the valuable part has been removed.

The practical question for anyone running global operations

The useful question for a FTSE leader isn't "should we still be offshoring." It's a sharper, more immediate one: is your current outsourcing partner becoming an AI-orchestration partner, or are they still billing you for seats? Ask them directly what proportion of their agents are AI-certified, whether their pricing model has actually changed to reflect a shift from headcount to outcomes, and what they've invested in reskilling compared to what they're still charging for volume. A provider still selling you the old model, at the old cost-per-contact logic, is a relationship that will be squeezed hard well before the next contract renewal, either by you or by a competitor already offering the new one.

The offshoring story was never really about the cheapest country to answer a phone in. It was always about who gets paid to know things a machine doesn't yet. That fight hasn't ended. It has simply moved indoors.