The Scorecard Nobody Wants to Audit

Now it's Heard 

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

This edition looks at the numbers we lean on and the ones we ignore.

Elvis Melia, researcher and former development practitioner, has spent 15 years watching the African GBS sector. He gives it the honest assessment it deserves and explains why the AI disruption story is more complicated than the headlines suggest.

Loud and Clear: Elvis Melia, questioning the data and asserting "not yet."

Back in 2020, Melia made the case that digital services could be Africa's development path, the way manufacturing was for East Asia. Six years later, with AI now in the conversation, he hasn't changed his position. But he is more honest about the uncertainty than most. 

"It's our best bet," he says. "The question is whether this time is really different." 

The familiar argument goes like this: new technology displaces certain tasks, creates others nobody anticipated, and we always find new things for humans to do. We went from fields to factories to offices. History says we figure it out. 

The counterargument is harder to dismiss than it used to be. There are serious economists who argue that this time, AI might be good enough to take on not just the jobs being displaced, but also the new ones being created. It has never happened before in history. Nobody knows what that world looks like, including the people building the technology. 

Melia sits with that uncertainty rather than papering over it. "I honestly don't know," he says. In a sector full of confident predictions, that is worth something. 

What the data does show, right now, is that the big BPOs have not shrunk. Concentrix and TP have roughly the same headcounts as three years ago. The shift of work offshore has been about cost, not automation. As Melia puts it: "As of late May 2026, AI is not yet capable of replacing human workers on net." 

That is not a prediction about tomorrow. It is an observation about today. 

The Honest Report: Actually, Quite Good 

Ask Melia for the real report on the African GBS sector and the answer surprises. 

"I think it's doing phenomenally well," he says. And he means it. 

To understand why, you need the numbers behind the numbers. In most Sub-Saharan African countries outside South Africa, around 85% of the workforce is in the informal economy. Of the roughly 15% formally employed, about half work for the state, in public institutions, where getting ahead is often more about who you know than what you can do. That leaves around 7-8% of workers in formal private-sector jobs. Here, the GBS sector is emerging as a critical growth engine. The rapid expansion of global giants into frontier hubs like Kenya and Ghana provides empirical proof that Africa can scale export-oriented services, where traditional export manufacturing has struggled to take root. 

"In that context, the BPO sector is like a superhero in the economy." 

People line up around the block for these roles. Not because they are the best jobs in the world, but because the real alternative is informal trading, or worse, going back to the village for subsistence farming. When workers take selfies in front of the buildings where they have just started their first formal job and send them home to their families, those photographs arenot naive. They are evidence of something real. 

He is also candid about why global BPO employers tend to treat workers reasonably well. It is self-interest. They want the “Great Place to Work” stamp. They worry about the journalist looking for a sweatshop story. "Whether or not they're nice people doesn't actually matter," he says, invoking Adam Smith. In this case, what is good for the company and what is good for the worker happen to point in the same direction. 

The 40% Automation Number: What It Actually Means 

A study by Caribou & Genesis Analytics, commissioned by the Mastercard Foundation, found that around 40% of GBS job tasks in Africa are already automatable with today's AI. 

Melia values this line of research and credits Jonathan Beardsley and colleagues. He is also clear about what that headline number actually means in practice. 

"We've seen figures like this before." In 2013, researchers at Oxford projected that 47% of American jobs were at high risk of automation within a decade or two. We are well into that second decade now, and the mass displacement has not materialized yet. But Melia is careful not to use that as proof that the researchers were wrong. 

The churn, not the displacement, is the real story. 

Does AI change that? Maybe. But it would be the first time in history. And nobody, including those closest to the technology, actually knows how it will play out. 

Are We Measuring the Right Things? 

The sector loves to count seats, headcount and revenue. Are those numbers actually telling us whether the work is delivering for workers and economies? 

Melia's answer is not the obvious one. 

"Revenue is the best metric," he says. Is what we are doing making money? Are people buying it? That is the clearest signal that real value is being created. 

But he also pushes back on the idea that headcount is simply an outdated number to be dropped in favor of outcome measures. His concern is practical. AI-generated work can look polished and complete while containing errors that are genuinely hard to catch until the damage is done: false citations, confident inaccuracies, and conclusions that sound right but are not. In that environment, having named humans in the process, people whose professional judgment is attached to the output, is a safeguard, not a relic. "It's a little bit like a seatbelt in the car." 

There is a second risk he takes seriously, one that gets less attention than hallucination. Sycophancy. AI systems that tell users what they want to hear, agree with bad ideas, reinforce rather than challenge. "Never ask an AI whether your startup idea is good," he says. "It will always tell you it's good." The messiness of human interaction, the colleague who disagrees with you, the client who pushes back: that friction is not a problem to be solved. It is part of what makes the work reliable. A setup that removes humans in the name of efficiency may be solving the wrong problem. 

One Day. Five Hours a Week. A Bigger Gap Than Anyone Admits. 

One of the more striking findings from Melia's research came out of a series of single-day AI training workshops run in Ghana and Rwanda with workers and trainees in the GBS sector. 

The format was practical. Participants brought a real problem from their working lives, wrote a detailed prompt, and ran it across seven AI tools at the same time: ChatGPT, Gemini, Claude, DeepSeek, Grok, and any two of Mistral, Meta AI, Perplexity or MS Copilot. Seven tabs, seven responses. They treated it like a job interview, narrowed the field to three, worked with those tools through the day, and finished with a three-minute pitch on the problem and the solution they had built together. 

Three months later, most participants reported that the AI workshop had changed their AI usage radically for the better: 69% reported that this one-day intervention now saved them at least 5 working hours per week. Over 80% felt “better” or “much better” at email writing, doing research, solving problems, creative work, and routine admin tasks. 

He is honest about the limitations of self-reported numbers. A separate METR study found that workers who thought they had become 20% more productive using AI were, by measured output, actually 20% less productive. Part of the reason: they handed work to the AI and spent the waiting time on social media, where they got stuck. 

But the finding Melia keeps coming back to is not the productivity number. It is what came just before it. When participants were asked what other AI training they had received before the workshop, around three-quarters said none, and 90% wanted more. 

This was in 2025. One day of AI training changed how people worked. 

"That tells you about the gap." 

What the Numbers Never Capture 

What does not show up in the headline data from Ghana and Rwanda are the things that matter most. 

The young workers who send photographs home tend to hold the first formal contract in their family's history. The shift, as Melia describes it, from a world built on who you know to one built on what you can do. "Is it the best salary ever? No. Is it better than anything else I could make? Yes." 

He also makes a point that gets little airtime in most sector conversations. We talk about digital natives, people who grew up with the internet. There is no equivalent term for AI yet, because generative AI has only existed since late 2022. Nobody grew up with it. 

Except, in a meaningful way, the young workforce, in countries where the median age is 18 or 19, has been using these tools in school, inside projects, in daily life, for the past three years. The same leapfrogging that happened with mobile money in Kenya, where a homegrown solution ended up being adopted globally, could happen again. A young workforce with no old habits to unlearn and real problems to solve is a different starting point than most people in the sector are accounting for. 

The Language We Use and What It Actually Means 

Cheryl Paarwater wrote recently for Heard about the words the sector uses: "bums in seats," "agent mentality," and how that language quietly shapes how workers are treated. Melia does not entirely disagree with her. But he uses the phrase himself, deliberately, with a specific argument behind it. 

"Those bums are inside pants that have pockets, and those pockets have space for more money." 

Money in pockets means braces for a niece. 

It means financing a mother's surgery. 

It means moving a family from subsistence to stability. 

The intention, he argues, matters more than the word. What he cares about is building real business relationships between African GBS operators and buyers in Europe and beyond, the kind of relationships that create trust, contracts and jobs for people who would otherwise not have them. 

On the language of AI specifically, his concern runs parallel to Cheryl's. Words like "disruption" and "transformation" mean one thing in a strategy session and something else entirely to a worker who was told this sector was their way to a career in the global knowledge economy and is now reading headlines about automation risk. 

"Quantify," he says. "Continuously quantify. Give a real picture of what AI adoption actually looks like on the ground today and take it from there." The boldest claims on AI disruption tomorrow, he notes, come from the labs that are simultaneously fundraising for their next most expensive models. That doesn’t mean they are lying, but it is worth keeping their incentives in mind. 

The Bottom Line 

The African GBS sector has survived every previous wave of automation anxiety without the mass job losses that were predicted. Whether AI is different this time is a genuinely open question, and Melia is one of the few people in this space who says so plainly rather than picking a side for the sake of a cleaner argument. 

What is not an open question is the training gap. One day of practical AI workshops drastically changed how 70% of participants worked. Most of them had received no AI training before that day. That is the real scorecard item nobody is talking about: not whether the seats are filled, but whether the people in them are being set up to stay there. 

Elvis Melia is a political economist and researcher specializing in African development, digital services and the future of work. 

His research on AI adoption in the African GBS sector, conducted for GIZ, is available on his Substack: The Impact of Artificial Intelligence. 

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.