From Call Center to AI Supplier: How Emerging Markets Can Reposition Their BPOs

For more than a decade, a call center job was one of the main ways for young people in South and Southeast Asia to get into formal office work. AI is now making that door smaller.

The numbers show it. Workforce analytics firm Revelio Labs reports that global call center employment is now 5.3% below its December 2023 peak. Revelio has tracked 70 quarters since 2009. Headcount fell compared with a year earlier in only eight of them, and those eight are the most recent eight quarters. Venture firm Andreessen Horowitz (a16z) has argued that AI is not yet causing wide job losses. Even so, its October 9, 2026 "This Week in Charts" newsletter singled out call centers as the exception.

Most people read this as bad news to manage: protect jobs and slow the decline. This article looks at it differently. The same change that is cutting call center hiring is also raising demand for what call centers already have: years of recorded customer conversations in local languages, agents who know what a good answer sounds like, and long-standing relationships with banks, telecom operators, and e-commerce platforms. Some BPOs are already earning money from AI work. Some AI companies are buying BPOs to get these assets.

Below, we look at how fast call center hiring is falling, why the business is still growing, how startups and BPO operators can make the switch, and what this could mean for jobs and export earnings.

Follow a market

The shift: call center hiring is falling everywhere

The Call Center Slowdown Began in Higher-Income Markets and Worked Down the Income Ladder

Change from a year earlier−12%+12%

In the 2010s, global call center headcount grew by about 4% a year. That trend has now reversed.

The decline began in high-income countries, where headcount started shrinking in 2022.

It then spread to lower-income countries.

By mid-2024, global headcount was shrinking overall. By Q4 2025, it was falling in every income group.

Revelio Labs, October 2026. Values read from the published chart's colours, so they are approximate.

To see why this matters, it helps to look at how the industry grew. For two decades, it was driven by cost. A call support hour in a high-income country costs about ten times as much as one in South Asia, so companies send the work offshore. That cost gap built the large BPO industries in the Philippines and India, and the smaller ones in Bangladesh and Pakistan.

High-Income Markets Pay Call Center Staff About 10 Times the Lower-Middle-Income Rate

Average pay per hour, by income group

Source: Revelio Labs, October 2026. Group averages.

AI changes that math. Buyers used to ask, "Where can I find the cheapest agent?" Now they also ask, "Does this call need a human agent at all?"

But is AI really the cause, or is the call center economy simply slowing down? The data points to AI. Revelio looked at 12 middle-income countries. Before ChatGPT launched in November 2022, call center jobs rose and fell in step with other office and administrative jobs. Since then, call center jobs have stalled or fallen, while other office jobs have kept growing. Revelio is careful here. It cannot see which firms replaced staff with chatbots, and it says several factors are at play. Still, it notes that this gap makes AI's role hard to dismiss.

Headcount index, November 2022 (ChatGPT release) = 100, latest reading

Revelio Labs, a16z (October 2026). Values are read from the chart and are approximate.

The decline is coming mostly from less hiring, not from layoffs. In the same 12 countries, the rate of people joining call center work fell from 25.1 to 17.2 per 100 workers since November 2022. The rate of people leaving fell from 22.3 to 18.5. Put simply, fewer young people are getting their first job in the industry. That raises a key question: what happens to today's workers, and to the young people who would have joined them?

Fewer People Are Joining Call Centers Than Leaving Them

Per 100 call center workers, 12 middle-income countries

Source: Revelio Labs, October 2026.

The paradox: fewer seats, more revenue

If call centers were simply dying, their revenue would be falling too. It is not. In the four markets that publish figures, industry revenue grew through 2025 and 2026. Over the same period, Revelio's data shows call center jobs stalling or shrinking.

IBPAP; NASSCOM Strategic Reviews 2025 and 2026; State Bank of Pakistan; BACCO.

How can both be true? The answer lies in what each data set measures. Revelio counts one job: the call center agent. Industry bodies count whole sectors, which now include healthcare and finance processing, IT services, AI data work, and global capability centers. The money is moving toward work that needs fewer, more skilled people for each dollar earned.

Industry leaders say this openly. In India, NASSCOM President Rajesh Nambiar described a non-linearity between revenue growth and job growth in FY2026. In other words, revenue is growing faster than jobs. In Pakistan, the Call Centers Association of Pakistan says AI tools have helped firms handle more work and win new international clients. Call center exports there rose 19% in FY2026.

The gap between countries also shows how much room there is to grow. The Philippine IT-BPM sector earns about US$21K in revenue per worker (US$40.3Bn across about 1.9Mn workers). Bangladesh's BPO sector earns about US$10K per worker (about US$1Bn across about 100,000 workers). These are author calculations based on industry figures and should be read as rough estimates.

The switch is already paying off

Some of the clearest evidence that call center businesses can change comes from companies that have already done it. They are selling the same workers' skills to AI companies, not only to customer service buyers.

TaskUs, a Nasdaq-listed BPO with large centers in the Philippines and India, now reports a separate AI Services business. This unit labels data, helps train AI models, and tests their outputs. AI Services revenue reached US$66.1Mn in Q2 2026, up 26% from a year earlier. It was the seventh quarter in a row of growth above 25%. At the same time, its Trust and Safety business shrank because its largest client automated that work. In other words, AI took work away from one part of the company, while another part grew by selling services to AI companies.

At TaskUs, AI Data Work Is Growing While Work Lost to Automation Shrinks

Year-over-year revenue growth, Q2 2026

TaskUs Q2 2026 results (Form 8-K filed with the U.S. SEC, August 2026) and earnings call

TP (formerly Teleperformance), which employs close to 500,000 people, is moving in the same direction. It reported strong momentum in its AI-powered services in Q1 2026, including data services for AI. Under its "Future Forward" plan, it will reinvest about 20% of its free cash flow into AI over 2026–2028.

Money is also flowing the other way. AI companies are buying BPOs to get their clients, agents, and know-how.

One AI-Native Buyer Has Closed a BPO Deal; Another Has a Ten-Target Pipeline

Crescendo bought PartnerHero, October 2024 · one dot = one employee, approximate

Crescendo at the deal, by its own figures

Crescendo's own figures, October 2024. Customers are for the company combined with PartnerHero.

Yellow.ai's Ten Targets Are Proposed, Not Completed

Estimated revenue per target, US$Mn. Five are in India. The dashed outline means proposed.

Sources: Crescendo and General Catalyst announcements; Yellow.ai and Bluerock investor presentation filed with the U.S. SEC. Target revenue is estimated from target management materials.

In October 2024, Crescendo, an AI-native customer experience company with about 20 employees, bought PartnerHero, a BPO with about 3,000 people. Crescendo was incubated by General Catalyst, a venture firm that has set aside US$1.5Bn for what it calls AI-enabled roll-ups: buying service businesses and running them with AI. General Catalyst says Crescendo's platform can fully automate more than 80% of customer interactions. PartnerHero's agents handle the rest, and they also train and check the AI. CEO Matt Price later called the deal "a bold move for a 20-person team to acquire a 3000+ person BPO" and said more BPO acquisitions are planned.

Others are following. Yellow.ai, a conversational AI company, has proposed buying ten BPO and customer experience firms, five of them in India, according to an investor presentation filed with the U.S. SEC. These deals have not been completed.

The trend is strong in India, where NASSCOM reports that 70% of the top 25 service providers bought AI-native firms in the past year. For this article, we found no completed purchase of a Bangladeshi, Pakistani, or Indonesian call center by an AI company. That means the local version of this model is still open to whoever builds it.

The playbook: from call center to AI supplier

AI agents can take over routine calls. But they cannot replace what call centers have built up over the years: data, judgment, and access to clients. AI companies need all three, and they are hard to find outside the region. Below are the three assets, the four business models they support, and the groups that stand to gain.

Three Assets Call Centers Already Hold Can Feed Four AI Businesses

What BPOs hold

What they can sell

Who gains

BPOs: A new revenue line
Agents: A path into AI work
The economy: AI services exports

Framework: ExitStack analysis.

Three assets

Conversation data. Many BPOs hold years of recorded customer calls in Bangla, Tagalog, Bahasa Indonesia, Hindi, and Urdu. These recordings capture real accents, the way people mix local languages with English, and the vocabulary of banking, telecom, and e-commerce. Languages such as Bangla and Urdu generally have little data available for training global speech and support models.

Experienced agents. An agent with several years of experience knows what a correct, polite, and complete answer sounds like in their market. AI developers pay for this judgment when they hire people to rate, rank, and correct what AI models produce. TaskUs's AI Services business is built on this kind of workforce.

Clients and operations. Local BPOs already work with the banks, telecom operators, and e-commerce platforms that will use voice AI. They also run secure offices, quality checks, and staff scheduling. An AI service needs all of these to run.

Four business models

1. Data licensing. With client approval, BPOs license call data to AI developers building local-language models. The data must be anonymized and collected with customer consent.

2. Annotation and evaluation services. Agents shift from answering calls to labeling data and rating AI outputs. This is the TaskUs model, applied to local languages.

3. Local-language voice AI products. Startups build voice and chat agents in Bangla, Bahasa Indonesia, Urdu, or Tagalog. They train and test them with BPO partners, then sell them to the same banks and telecom operators.

4. AI-native BPO. A startup buys or partners with a BPO. AI handles routine calls, while agents handle complex cases and supervise the AI. Clients pay for results rather than hours. This is the Crescendo model.

Major investors are betting on this model. In a March 2026 essay titled "Services: The New Software," Julien Bek of Sequoia Capital argued that the next US$1Tn company will be "a software company masquerading as a services firm." His reasoning is simple. For every dollar companies spend on software, they spend six on services. As Bek puts it, "A copilot sells the tool. An autopilot sells the work." He also points to outsourced tasks as the easiest place to start, because buyers already pay outside firms to do them. That is the work BPOs do today, and it is the same bet General Catalyst made with Crescendo.

For startups, a practical route is to partner with a BPO whose margins are under pressure. The startup licenses the BPO's data, hires its agents to label and review AI outputs, and sells through its client base. Each side gains. The BPO gets a new revenue line. Agents get a path into AI work. The startup gets local-language data that is hard to collect from abroad.

The economic case

This matters beyond any single company. Call centers have been a main way into office work for young people across the region, including many women. In Bangladesh, BACCO reports that about 40% of the sector's workers are women. The ILO has warned that entry-level office jobs like these are the most exposed to automation in lower-income countries. The shift affects three groups: workers, exporters, and the wider region.

For workers: a ladder instead of a sideways step

Revelio's data shows the risk clearly. Better-paid jobs in these same economies are growing, but most people who leave call centers do not move into them.

One square = one of 100 people finding a job within 12 months of leaving a call center

The Growing Roles Pay More Than Call Center Work

Pay premium over call center work; jobs growth over the last year

Pooled across 12 middle-income countries. No figure for a single market.

Revelio Labs, October 2026, Figures 5 and 6; 'other destinations' is the remainder

AI data work, such as labeling, evaluation, and AI supervision, could close that gap. It sits between the call center floor and those better-paid jobs. It draws on the judgment agents already have and builds the technical skills the growing roles need. If BPOs and startups build this bridge on purpose, more call center workers could move up instead of sideways.

For export earnings: more revenue per worker

Bangladesh's BPO sector earns about half as much per worker as the Philippine IT-BPM sector. Part of that gap comes from the Philippines' focus on exports and on higher-value work such as healthcare and finance processing. Closing even part of the gap does not require more seats. It requires higher-value work from the seats that already exist.

Author calculation from BACCO (about US$1Bn, about 100,000 workers) and IBPAP (US$40.3Bn, about 1.9Mn workers, 2025)

AI data and evaluation work is one way to get there. It is sold to global buyers and paid in U.S. dollars. It also depends less on English than voice support does, because the local language is the product. This matters for Bangladesh, Indonesia, and Pakistan. Their BPO sectors are smaller, and in Bangladesh and Indonesia, they mostly serve local clients.

For the region: demand is already shifting online

According to the World Bank's Working Without Borders report, there are an estimated 154Mn–435Mn online gig workers worldwide. Demand for online gig work grew 41% between 2016 and Q1 2023, and it grew faster in developing countries. Low- and middle-income countries already account for 40% of traffic to online work platforms. But much of this work, including data labeling, runs through informal platforms with little job security. Routing it through established BPOs, with formal jobs and benefits, is one way to capture this demand. It would also protect the job quality that made call center work a step into the middle class.

What has to go right

The opportunity is real, but it will not happen on its own. Four conditions will decide whether the region benefits.

Data rights and consent. Call recordings usually belong to the BPO's clients, not the BPO. Their use is governed by client contracts, customer consent, and local data protection rules. Any licensing deal needs client approval, and the data must be anonymized before it is used for training.
Wages and job quality. Data labeling on informal platforms has been criticized for low and unstable pay. The economic case only works if this work is formal, salaried, and paid at least as well as call center jobs.
Steady demand. TaskUs notes that AI training and safety contracts are often one-off projects rather than recurring work. BPOs entering this space need a steady pipeline of clients, not a single large contract.
Timing. As automation spreads, there will be fewer calls handled by people and fewer experienced agents. These assets are worth the most now, while the expertise is still in the building.

Conclusion

AI is taking routine calls away from call centers in Bangladesh, the Philippines, India, Indonesia, and Pakistan, and hiring data now shows it. Yet revenue is still rising in the markets that report it. Companies like TaskUs are growing AI services businesses with the same workforce. And AI companies are buying BPOs for what they hold.

So the real choice for emerging markets is not between saving call center jobs and losing them. It is about who captures the value. Local BPOs and startups can turn their data, agents, and client relationships into an AI supply industry. Or others can buy these assets and take the value elsewhere. No AI company has yet bought a call center in Bangladesh, Pakistan, or Indonesia. That leaves room for local founders, BPO operators, and investors to build the model first.