The business process outsourcing (BPO) industry is no longer constrained by technology limitations. Artificial intelligence, automation and agentic systems are already capable of handling large volumes of repeatable work at speed and scale. Yet the transition to AI-led outsourcing is progressing far more slowly.
According to a new Market Impact Report by HFS Research, produced in partnership with business process management company Firstsource Solutions, it’s not AI maturity that is holding up transformation, but the way BPO services are contracted, priced, and governed.
The report, The Future of BPO: Services-as-Software, notes that the BPO industry is at a “tipping point” as enterprise needs don’t necessarily align with their headcount-based delivery, fixed pricing, and labour-intensive operating models. In a digital-first, AI-enabled economy, buyers are prioritising speed, transparency, resilience, and measurable outcomes over offshore delivery centres and full-time employee (FTE) counts.
From Labour to Outcomes
The limitations of the traditional FTE model are well understood. Contracts structured around effort and headcount incentivise providers to sustain work rather than eliminate it, leaving enterprises paying for labour instead of value. Rigid scope, slow change management, and siloed accountability further restrict the ability to adapt services as business conditions evolve.
HFS survey data suggests buyers are ready to move on. Reliance on FTE-based commercial models is expected to fall from 42% today to 28% over the next three years, while next-generation BPO models—characterised by outcome-based and shared-revenue pricing—are projected to rise from 20% to 39%.
But while buyer intent is clear, execution remains uneven.
Where AI-led BPO is Stalling
The report explicitly states that AI availability is not the binding constraint. Instead, adoption is being dragged by organisational inertia—particularly across procurement, finance, and governance.
Enterprise leaders interviewed by HFS say procurement teams remain wedded to predictable, FTE-led pricing and resist outcome-based or subscription models, even when automation fundamentally alters delivery economics. Flat-rate pricing, favoured for cost certainty, clashes with AI-driven services that improve continuously and scale non-linearly.
Data readiness is another structural obstacle. Clean, integrated and governed data remains elusive for many enterprises, limiting the ability to deploy AI-led services at scale. Security, auditability, and explainability further complicate adoption, especially in regulated environments.
Operational realities reinforce this caution. Resistance to automation has historically stemmed from revenue cannibalisation fears in headcount-led contracts.
“Most of the time, it was always observed that the revenue cannibalisation will happen because of RPA (robotic process automation) or the revenue cannibalisation will be severe if we kind of deploy any AI programs,” Naushath Raja Mohammed, Head of Artificial Intelligence and Machine Learning CoE at Sutherland, tells AIM.
That mindset, he adds, is now shifting as providers reposition around AI-led delivery. “From a BPO standpoint, we are transforming our business models… BPaaS model… human plus AI plus digitisation… organisation culture is always AI first.”
The Delivery-Risk Question
Beyond governance and data, a more fundamental issue is emerging: where does delivery risk actually sit in outcome-based, AI-led contracts?
HFS argues that outcome-based pricing cannot simply be layered onto legacy delivery models. It only works when deep domain expertise, end-to-end process ownership and technology control operate together. Without that triangulation, outcomes become unpredictable—and delivery risk unmanageable.
Hasit Trivedi, Chief Digital and AI Officer at Firstsource, agrees, saying that the shift is already visible in live contracts.
“Outcome-based pricing is not a commercial construct that can be layered on top of legacy delivery. It only works when it is backed by deep domain expertise, process ownership, and technology control working together,” he notes.
“When domain, process, and technology come together effectively, outcomes can be delivered predictably. When they do not, outcomes can be impacted, and in such cases, our topline and bottomline are impacted as well, because in outcome-based deals the provider takes on a larger share of delivery risk.”
“Services-as-software does not eliminate delivery risk; it redistributes it to where it can be best managed,” he adds.
BPO Durability
The redistribution of risk is also reshaping how large BPO providers position AI-led operations. Leaders argue that productivity gains from automation are increasingly being shared with clients, while providers take on greater accountability for execution and outcomes.
Speaking during Cognizant’s earnings call, CEO Ravi Kumar said the company is structuring BPO engagements around productivity sharing and outcome ownership.
“If you look at it, in some ways, we are sharing the productivity, sharing the risk with our clients, but we are actually doubling down on execution,” Kumar said. “Look at our revenue per person and margin per person… it has gone up… which essentially means we are able to share with our clients the productivity win.”
He added that AI and platformisation are enabling BPO providers to move back toward owning operational outcomes rather than billing purely for effort.
“Historically… companies like ours used to own the outcomes, and we used to price on outcomes… Here we are fast-forward. We can own the outcomes. We can make this a platform play… nonlinear cost and nonlinear revenue.”
Kumar also framed BPO as a structurally expanding market in the AI era, driven by the scale of enterprise operations and the need to run and maintain AI-enabled processes.
“This is a total addressable spend… embedding technology data into process… we think this is a longish tailwind,” he said, adding that maintaining probabilistic, AI-driven systems could increase operational demand rather than reduce it.
The Credibility Gap
Even where enterprises are willing to experiment, HFS identifies a growing credibility gap between what providers promise and what they can deliver in production.
Buyers told researchers that many vendors promote modular, AI-led services but lack evidence that these models work at scale. As a result, the intention to move toward services-as-software often outpaces actual implementation.
Sceptical buyers are increasingly looking beyond vendor-controlled case studies. Independent analyst validation, platform-level transparency metrics such as uptime and MTTR, third-party certifications, and sustained financial performance are emerging as key proof points.
Trivedi points to multiple signals enterprises now rely on to assess execution credibility, including hyperscaler partnerships, analyst positioning across digital operations and AI, and performance transparency embedded into platform-led models.
“Services-as-software is no longer theoretical,” he declares. “It is being validated through ecosystem behaviour, independent assessment, operational transparency, and sustained performance. Buyers don’t need to accept that on faith; they can see it in how the market is already moving.”
Phased Transition
The report stops short of predicting an abrupt end to traditional BPO. Instead, it outlines a phased transition, where AI-led and platform-based services are progressively integrated into existing delivery models. Human oversight will remain critical, particularly in regulated industries and high-value customer interactions.
Operationally, many enterprises are starting small—piloting AI in targeted workflows, validating security and data integrity, and scaling only after measurable gains are established.
As Mohammed notes, the practical approach remains iterative. “Start small… see security… scale fast,” he remarks, emphasising that data quality, environmental complexity and straight-through processing capability (automating payments and securities trade transactions automatically) determine whether AI deployments succeed in production.
Your reaction
Discussion
What's your take on this story?
No comments yet — be the first to share your view.