The Market Is Writing BPO’s Obituary. Is the Industry Listening?

Now it's Heard 

The industry has a voice. Now it’s Heard.

We launched Heard to surface perspectives that challenge how the BPO sector thinks and operates. This edition turns to a question the industry keeps deferring: not whether the model needs to change, but who will own the risk when it does.

Andrew Wrobel, Founder and Chief Reinvention Officer at Reinvantage, argues that BPO's reinvention is not a storytelling problem. The model that built the sector, selling hours against headcount, is being quietly undermined by the same automation providers that are being asked to deliver. His case is that expertise can either defend what currently pays or be redirected toward understanding why it is losing relevance. That choice, he writes, is the same one facing the fund manager in Warsaw who inspired the piece.

Loud and Clear: Reinvent or die

Outsourcing grew rich by selling hours. Its reinvention depends on changing the way it values work.

At a recent Fund Forum in Poland, I listened to two very different readings of the same asset-management market. One fellow panellist defended the established industry: funds should be boring, portfolio construction belongs to professionals, and newer forms of investing are too often driven by speculation or inexperience. Another argued that the industry had to accept that access, technology, and customer behavior had changed, and that expertise could not become a reason for dismissing what clients were choosing instead.

The disagreement was not about whether expertise still mattered. It was about what that expertise was being used for: to defend an existing model, or to understand why it was losing relevance.

That argument reminded me of Adam Galinsky, the Paul Calello Professor of Leadership and Ethics at Columbia Business School, who offers a useful frame for why this situation is hard to see from the inside. In one experiment, participants asked to recall a moment of power were nearly three times more likely, when asked to draw a capital E on their own forehead, to draw it from their own perspective, backwards to anyone facing them. The more firmly you occupy a position of authority, the easier it becomes to mistake your view of the system for the whole.

This is not just typical of individuals: industries do this too. Entire sectors can begin to interpret every new challenge in the vocabulary they already have. For BPO, that vocabulary is processes, locations, FTEs, service levels, etc.

A recent post by Mark Hillary compared BPO to a wooden ship on fire, with short sellers, analysts, and AI vendors increasingly defining its future while many of its largest companies struggle to explain what they're becoming. Hillary is right that the sector cannot allow everybody else to write its obituary, but BPO's problem runs deeper than its inability to tell a compelling story.

Bums on seats

The sector's founding proposition was simple and remarkably successful. Companies took activities they already performed, such as answering enquiries, processing invoices, administering claims, managing payroll, and reconciling accounts, and transferred them to specialist providers that could perform them more consistently and, above all, cheaper. The provider's job was to absorb the activity, standardize it, move it somewhere more efficient, recruit the people, and meet the agreed service levels. It was rarely rewarded for asking whether the work should exist at all.

This led to more work requiring more people; more customer contacts meaning more agents. Growth came through additional seats, hours, transactions and processes. The language has since evolved from outsourcing to partnership, transformation and value creation, but the underlying arithmetic has stayed tied to labor, and most markets still measure the sector chiefly by the number of jobs it creates.

I saw how deeply embedded that logic remains in my own company, less than two months ago. We asked a provider to help us design an outcome-oriented project. We described the results we wanted rather than prescribing the activities, team structure, or number of people involved. The response matched our outcomes to job descriptions and offered a selection of FTE configurations. We had asked what it would take to deliver the outcomes; the provider answered with how many people it could assign.

That isn't a failure of competence, but evidence of how the sector has learnt to price value. People are easy to count, and their time can be priced, allocated, and written into a contract. Outcomes are harder: they require the provider to make assumptions, exercise judgement, share risk and accept greater responsibility for whether the work actually produces the intended result.

For the buyer, an outcome-oriented model sounds logical. For a provider built around headcount, it can be commercially destabilizing: the provider is no longer being asked to supply a team and perform an agreed set of activities, but to determine which activities are necessary, which should be automated, which should disappear, and to commit to a result without controlling every condition affecting it.

Time, please

This is where AI changes the question, not just the economics. The old question was: Where can this work be done more cheaply? The new one is: Why is this work being done this way at all? A customer contact that can be prevented doesn't need to be handled more cheaply. An invoice moving automatically through a well-designed system doesn't need to be processed in another location. A claims process built on autonomous verification may not need the same chain of manual checks. A report produced continuously by an intelligent system doesn't need to be assembled at the end of every month. If revenue still depends on FTEs, automation improves delivery while quietly weakening the model that pays for it: the provider is asked to remove work operationally while still depending on it economically.

That tension points towards hybrid pricing: platform or capability fees, transaction-based pricing, committed capacity where it remains genuinely necessary, agreed service and experience outcomes, and shared gains where measurable improvements are achieved.

But hybrid pricing only survives if it is incentive-compatible for the provider, not just logical for the buyer. A provider currently protected by FTE billing has no commercial reason to move towards shared-gain pricing unless the upside from designing a better system exceeds the revenue it gives up by removing work. Reinvention here isn't a pricing preference but a bet that owning the outcome is worth more than owning the headcount.

Other professional-services industries face a version of the same problem, and my own strategic reinvention firm is not excluded. Management consultancies have traditionally charged for time, expertise and capacity. AI may let us finish parts of our work faster, but that efficiency raises an uncomfortable question: Why should a client keep paying for hours that technology has removed? For BPO the tension is more fundamental, because so much more of its revenue was built on exactly those hours.

Public markets appear to be already drawing a line inside the sector. Labour-intensive customer experience names sit firmly in the AI-risk bucket: Teleperformance has become one of Europe's most-shorted stocks, with one fund manager telling the Financial Times that call center work amounts to "selling human time to handle repetitive tasks," and Barron's reports Concentrix down more than 52 percent over the past twelve months. TELUS Digital fell more than 24 percent before its parent moved to take it private at a premium to the depressed price. Accenture (far broader than pure BPO but caught in the same repricing) fell to its lowest level since 2017 in June, an 18 percent single-day drop on weak guidance that took its market value from over $200 billion at its post-Covid peak to under $80 billion.

What gets rewarded runs across that same line. Infosys and TCS have both seen their shares rise when AI-led demand shows up in stronger guidance and order books; investors respond to evidence that AI is driving revenue, not just cutting costs. Cognizant has had similar moments but has also lost more than a third of its own market value this year, weighed down by the same AI-driven deflation fears hitting Concentrix and Teleperformance. The line the market is drawing isn't about geography or branding. It appears to be about whether a business is still selling labor against repetitive tasks or can show a credible route to AI-led revenue instead.

Some service providers have already been testing that bet, and it's worth noting where. Independent reporting in the Wall Street Journal described EXL, a provider built on outsourced back-office and digital operations, as cutting or redeploying hundreds of roles while recruiting for more advanced data and AI skills, with analytics then accounting for roughly 45 percent of revenue.

That's a single snapshot rather than an ongoing ratio, but later coverage confirms the direction rather than reversing it: by January 2025, EXL had launched an insurance-specific large language model built with Nvidia and was recording double-digit growth in analytics and digital operations, and a mid-2025 interview described the company continuing to redesign workflows around data, AI, and industry expertise, though through company-supplied examples. None of that proves EXL has escaped labor-based economics; a contract can still be billed by the hour even when those hours go on analytics rather than call handling. But it shows a sustained capability shift, not a single announcement.

Outcomes and incomes

Genpact makes a related but sharper case. It began as GE's captive BPO operation, processing loans and credit-card transactions, before becoming independent. Its 'Client Zero' program tests AI-enabled operating models inside Genpact's own business before offering them to clients. The company reports close to $40 million removed from its own operating expenses this way, with some invoice-processing cycles cut from weeks to hours, and those tools are now being piloted with clients. Those figures are company-reported, with no independent source confirming them and no public disclosure of how the resulting client contracts are priced. What the case still demonstrates is the posture: Genpact is exposing its own delivery base to the automation it's trying to sell, rather than treating that base as something to protect.

WNS remains the strongest commercial signal of the three, and now on a longer evidence trail than the other two. It began in 1996 as a British Airways back-office operation in Mumbai, becoming independent in 2002. Capgemini's $3.3 billion acquisition of it in 2025 was framed explicitly as a shift from business-process services to agentic-AI-powered intelligent operations, with Capgemini's own announcement tying that shift to commercial form, citing transaction-based, subscription-based and outcome-based pricing as the opportunity AI creates.

By February 2026, Capgemini reported around 100 WNS-related cross-selling opportunities and a €600 million intelligent-operations contract spanning multiple functions. By April 2026, Reuters was reporting that Capgemini's first-quarter revenue had grown 7 percent at constant exchange rates, with North American revenue up 20.7 percent partly on the strength of the newly integrated WNS business, and generative and agentic AI still above 10 percent of group bookings. None of this discloses how the €600 million contract is actually priced. But it's no longer just an acquisition announcement: it's a business that's been folded in and is measurably contributing to growth three quarters later.

So, across the three: EXL shows the talent and revenue mix moving first. Genpact shows a provider willing to automate its own work before selling that automation. WNS/Capgemini shows the commercial conversation naming non-linear pricing explicitly, backed by a very large contract, though not by disclosed contract terms.

Together, these are established BPO providers (not just companies adjacent to the industry) changing capability, workforce and commercial direction, which demands more than simply redescribing what they already do.

The cost of making the wrong bet is not hypothetical. Capital is rotating hard towards the infrastructure layer of AI itself: the FT reports that Alphabet, Amazon, Meta and Microsoft are on track to spend about $725 billion on AI infrastructure in 2026. Credit markets are also becoming more selective. JPMorgan has warned that concentrated, long-dated AI-related debt issuance can pressure technology spreads wider. Reuters has reported that JPMorgan marked down some software-exposed private-credit loans, Blue Owl plans to reduce software exposure in one of its private-credit funds, and HSBC is pulling back from riskier private-credit lending.

The exact leverage reset inside CX and BPO deals is harder to document publicly, but the broader direction is clear: investors and lenders are giving less benefit of the doubt to cash flows seen as vulnerable to AI-led disruption.

Capita is the sharpest illustration inside the sector itself. Its 2025 annual-report materials show revenue down 4.5 per cent, a pre-tax loss of £170.9 million, free cash flow of negative £82.1 million, and net debt of £461.6 million. By May 2026, the company's market value had fallen to about £365.5 million, below that year-end net debt. In July 2026, failures on the civil-service pensions contract triggered withheld payments, a new profit warning and a further delay to the target for positive free cash flow, now pushed back to 2027.

That is not proof of AI disruption alone. It is evidence of eroded equity value and shrinking financial flexibility when weak growth, thin cash generation and an ageing commercial model compound faster than a business can redesign around them.

Nobody has yet publicly demonstrated the commercial success of outcome-based pricing at scale in BPO; that remains the honest limit of this argument. What the public record does show is which bet established providers are choosing to place, and what it costs to avoid placing one. EXL, Genpact and WNS/Capgemini are on one side of that line. Capita is on the other.

Which brings me back to Poland. My fellow panellist's expertise was being used to explain why the existing model remained right. The other position was that expertise couldn't be a reason to dismiss what clients were already choosing instead. BPO faces the identical fork: expertise can defend the delivery and pricing model that currently rewards the provider, or it can be redirected toward understanding why the buyer's world has changed. It's the same the same asset, same choice, different industry.

BPO was built to do work buyers already did, only more cheaply. Its reinvention depends on helping buyers decide what work should exist at all, and on being willing to own the outcome once the busywork is gone.

Follow Heard

Get the full picture every week: Subscribe to our email newsletter for weekly signal tracking across the industry.

Be part of the conversation: Follow BPO Bullhorn on LinkedIn for debates, polls, and industry insights.

Don’t miss the next conversation that’s shaking up BPO and CX. Tap Subscribe on our LinkedIn newsletter to get future editions of Heard straight to your inbox.