AI is Changing the Way Outsourcing is Bought, Sold, and Priced
For three decades, business process outsourcing (BPO) ran on a reassuringly simple equation: more work meant more people, preferably somewhere cheaper.
AI is dismantling that equation, but not necessarily the industry built on it.
The evidence across operators, their customers and the markets points in the same direction. Companies still want more work done, but they increasingly expect output to rise without headcount moving in lockstep. That is turning an industry built around selling people and hours into one that will increasingly be judged on productivity, automation, judgement, and outcomes.
Tech research firm Information Services Group (ISG)’s 2026 Global State of BPO Report, which is based on a survey of 250 large-enterprise buyers, makes the disconnect explicit. 70% of customers expect their providers to deliver more innovation and a broader range of services. Only 40% expect the number of outsourced workers to increase.
Same demand. Fewer incremental humans.
Deal data suggests some of that pressure is already appearing commercially. ISG recorded US$27.3bn of BPO contract value in 2025, down from $30.4bn in 2019, while average award value fell from $14.1m in 2023 to $11.8m in 2025.
The industry is being asked to do more with less, and the "less" is increasingly headcount.
AI Deflation Arrives
Equity analysts have started calling the underlying mechanic "AI deflation". Providers become more productive, but instead of allowing those gains to drop neatly into their own margins, customers demand that at least some of the saving comes back through price.
That is already visible in major outsourcing relationships. Reuters reported in August that TCS, Infosys, Wipro, HCLTech and Cognizant are all shifting parts of their commercial models away from hours worked and toward outputs and outcomes. TCS has said roughly 80% of its finance, HR and business services contracts now use outcome measures. Cognizant has gone one step further in its Daimler Truck relationship by explicitly sharing AI-driven savings between customer and supplier.
That is a very different proposition from the traditional model in which revenue growth was closely linked to hiring growth.
The more important distinction now is not between outsourced and insourced work. It is between tasks that can be automated, tasks that can be augmented and tasks that still require human judgement.
Routine, high-volume work with digital inputs and clear quality standards is the obvious first target: invoice matching, password resets, claims intake, basic reporting, templated customer support, and simple reconciliation.
At the other end of the spectrum sit activities involving domain knowledge, regulatory interpretation, exceptions, complex problem solving, or accountability. Those are much harder to automate fully.
ISG's own data reflects the divide. Industry-specific BPO, where judgement and specialist knowledge matter more, has grown bookings by 27% since 2019. Customer experience and back-office work remain below 2019 levels.
The lesson is not that functions disappear. Tasks inside them do.
AI Plus Offshoring, Not AI Versus Offshoring
One of the most important insights from our conversations with technology companies is that AI does not necessarily cause outsourced work to come home. In some cases, quite the opposite is happening.
One technology company we interviewed is consolidating engineering from several Asian locations into India at the same time as it uses AI to increase speed to market and reduce costs. This is important because much of the public debate assumes a binary choice: either AI automates the task, or the task remains offshore.
The reality is more complicated. A company might automate 20% of a process, move another 30% into the Philippines or India, redesign the workflow and still end up with a much smaller total workforce. AI and offshoring can be complementary for a considerable period before the full headcount effect becomes visible.
That is why the strongest offshore hubs may actually gain higher-value work during the transition.
Providers Are Not Standing Still
For any outsourcing provider dependent primarily on staffing scale, this is clearly uncomfortable.
But not every incumbent sees AI as a threat. Cognizant CFO Jatin Dalal has argued that generative AI is expanding the addressable opportunity in parts of its BPO business. He points to combinations of people and AI agents that can make smaller and more dispersed delivery operations economical without requiring the traditional model of concentrating hundreds or thousands of workers in one offshore centre.
That points toward a different delivery architecture. AI handles standard execution. Humans handle exceptions, judgement, and accountability. Providers contribute workflow knowledge, systems integration, governance and increasingly the orchestration of multiple AI agents.
The economic value shifts away from simply supplying capacity. The winners will be the providers able to combine proprietary process knowledge, technology integration, and real operational accountability. The losers will be those still selling headcount as though nothing has changed.
Outcome Pricing Is Coming, but Not Overnight
The logical commercial endpoint is outcome-based pricing. Instead of paying by employee, hour or seat, customers pay for processing time, claims leakage, customer satisfaction, collection rates, completed transactions, or some other measurable result. That sounds simple until one tries to do it.
Our interviews suggest the obstacle is often operational rather than philosophical. One technology company we interviewed is progressively moving toward outcome-based pricing, but only after installing tools that allow output to be measured at employee and process level. Another expects the same transition but believes it will take several years. The lesson is wonderfully mundane: before you price an outcome, you need to be able to measure it.
Attribution makes matters harder still. If a process becomes 40% cheaper because a business introduces AI, moves work offshore, redesigns the workflow and changes staffing at the same time, who gets the credit? More importantly, who gets the saving?
That question will sit at the centre of contract renegotiations over the next several years.
Clients Have Leverage, For Now
Buyers therefore have a rare opportunity. They should be asking providers to demonstrate that transaction volumes can rise without staffing rising alongside them. They should also challenge long-term commercial models still linked almost entirely to FTE growth. And they should demand evidence of AI capability beyond PowerPoint slides and pilot programmes.
But clients should not leap blindly from time-and-materials contracts into full outcome pricing, either. The better route is a glide path. Establish the current baseline. Measure productivity. Build reliable attribution. Share gains during the transition. Then migrate progressively toward outcomes where the underlying data is good enough to support it.
The practical mistake would be to sign a five-year agreement whose economics assume that every extra unit of demand requires another person.
The New Outsourcing Equation
Outsourcing demand has not disappeared. What is disappearing is the assumption that more outsourced work requires more outsourced people.
During the transition, the strongest offshore centres may gain activity as companies consolidate work into fewer locations and move those centres up the value chain. Eventually, however, they face the same arithmetic as everybody else: more output, fewer people and increasing pressure to price what gets achieved rather than how many hours were spent achieving it.
AI is not killing outsourcing - it is forcing outsourcing to become something else.
See also: AI's outsourcing shock exposes a tale of two markets in India and the Philippines.