The market is understandably focused on the headwinds that artificial intelligence creates for the services industry. If AI can automate work, improve productivity, and reduce the labor required to deliver services, it follows that outsourcing providers should eventually face significant revenue pressure.

However, that is only one side of the story. AI is also creating tailwinds, and one of the clearest examples is in business process outsourcing, or BPO. In fact, the expectation that AI should deliver dramatic cost savings is, somewhat counterintuitively, helping drive more work into traditional BPO models today. What makes this particularly important is that it challenges the prevailing assumption about how AI disruption will unfold in services. The long-term direction may still be toward greater automation and lower labor intensity, but the path from here to there is not linear. In the near term, AI is creating new pressure on companies to cut costs, expand productivity, and rethink how work gets done. Those pressures are producing some unexpected outcomes, including more demand for established outsourcing and offshore delivery models. To understand why, it helps to look first at where growth in BPO is actually coming from.

Where BPO Growth is Coming From

Growth in services generally comes from three places; transaction volumes inside existing contracts increase; providers sell additional scope to existing customers; or alternatively, they sign up new customers.

All three are occurring in BPO, but much of the above-normal growth we are seeing is coming from additional scope. Companies such as EXL, Genpact, and others have been performing above both broader market growth and, in some cases, their own historic norms. Why is this happening?

AI is Adding Fuel to the Fire

AI is adding momentum to this trend. There is an extraordinary expectation now being placed on AI-driven productivity. I have heard several instances of CEOs meeting with frontier AI companies, returning to their organizations, and instructing business functions to find savings of 40%, but there is a problem with that mandate. For most functions, AI cannot currently produce savings of that magnitude simply by being added to the way work is done today. Instead, the way we approach the work needs to be reinvented, and I think it is important to separate the two very different paths AI implementation can take. The first is evolutionary; companies apply AI to existing processes to make people and functions more productive. This is already happening, and it can produce meaningful benefits. The second is transformational; companies build native agentic platforms and radically reinvent how a business function operates.

The very large productivity claims associated with AI generally depend on this second path. If a company genuinely wants savings in the 20% to 40% range, it will usually need to redesign processes, operating models, workflows, and organizational structures around an AI-native or agentic environment, but this does not happen quickly.

These platforms need to be built, tested, governed, and proven. Companies also need confidence that the redesigned process will work reliably because the cost of getting a major business process wrong can substantially exceed the benefit of getting it right.

As a result, the transformational path is measured in years. Unfortunately, many corporate cost mandates are measured in quarters.

Why these High AI Expectations are Helping BPO Providers

This timing mismatch creates the first BPO tailwind. When a business function is told to deliver substantial savings quickly but cannot achieve them through AI alone, it turns to proven cost-reduction mechanisms. Labor arbitrage and offshore delivery remain among the fastest and most predictable options available. BPO does not necessarily produce the full 40% saving a CEO may have requested, but it can move the organization meaningfully toward that target. Consequently, some companies are expanding the scope they give to BPO providers precisely because they have been asked to capture the savings associated with AI.

The second tailwind comes from providers themselves. BPO companies are layering more AI, automation, analytics, and related services onto their existing work. As they do so, they often expand into adjacent activities and take on additional scope.

These two effects are reinforcing each other. Customers are looking for additional savings, while providers are using AI to broaden what they can offer. So far, the resulting growth has more than offset the revenue compression that AI might eventually create through automation.

It’s Not Unique: The Same Pattern is Emerging in GCCs

We can see a similar dynamic in global capability centers, or GCCs. Companies continue to expand their GCCs and move additional business functions into offshore captive operations. One reason is again the demand for cost reduction. If leadership asks a function to produce substantial savings quickly, moving more work into an established GCC is one of the simplest and most reliable options available.

This is particularly interesting because AI was expected by some to make these traditional cost-reduction models less relevant. In the near term, it is doing the opposite. The expectation of AI-driven savings is actually accelerating the use of mechanisms companies already know can reduce costs.

The Near-term and Long-term Effects of AI and its Impact

There is a broader lesson here. The benefits of AI are likely to arrive much more slowly than the conventional narrative suggests. That does not mean AI will fail to transform businesses. It means transformation is constrained by organizational reality.

Change takes time. Process reinvention is painful. Companies need to experiment, test, manage risk, and ensure that each step does not damage the operations they are trying to improve. This is also why I believe it is a mistake to frame AI strategy as a choice between evolution and reinvention - companies will do both.

It makes no sense to stop improving today’s processes while waiting for an entirely new agentic operating model to mature. At the same time, companies will increasingly begin building agentic platforms within individual business functions.

These environments can start small and coexist with existing processes for many years. In some areas they may eventually replace the old operating model. In others, they may continue alongside it because the agentic platform creates new services and capabilities rather than simply replacing what already exists.

The market tends to focus on AI’s threat to the services industry because the long-term automation story is easy to understand. However, the near-term economics are more complicated. Expectations for AI savings are rising much faster than companies’ ability to deliver those savings through AI alone. That gap is helping drive additional work into BPO providers and GCCs, while also allowing providers to expand their scope through AI-enabled services.

Eventually, agentic AI will transform many business functions, but the transition is unlikely to be a multi-year event. I believe we are looking at a multi- decade process in which companies simultaneously improve the operating models they already have and gradually build the ones that may ultimately replace them.