One Client to Rule (and Kill) Them All

Waking up to news like the Sama layoffs in Kenya feels like watching a server rack catch fire in real-time. 1,108 people. That’s not just a number on a spreadsheet; that’s over a thousand desks suddenly empty because one giant in Menlo Park decided to flip a switch.

As a dev, I’m constantly thinking about "single points of failure." We lose sleep over a single API that might go down or a database without a replica. But in the world of business building, especially in Africa, the biggest single point of failure is often your client list.

The Anchor that Became a Sinking Weight

Sama was basically the "impact" darling of Nairobi. They had the big names—Google, Microsoft, Ford—but everyone knew Meta was the one keeping the lights on. It’s the classic trap. You get a massive contract from a Big Tech player, and suddenly your whole operation is optimized for their specific workflow. You’re not building a diversified business anymore; you’re building a specialized department for someone else’s company.

In the software world, we call this "vendor lock-in," but this is the human version of it. When Meta decided to pivot toward automated filtering and cheaper providers, Sama didn't just lose a client; they lost their foundation. It reminds me of the guys I know in Akure who spend all their time white-labeling for one US agency. It’s great money until the agency finds a cheaper dev in Southeast Asia or decides to "optimize" with AI. Then, suddenly, Sapa starts knocking.

The Automation Trap

The irony here is thicker than the humidity in Gbagada. These folks were hired to label data to make AI smarter. They were literally training their replacements. Meta is now saying they can do more with automated filtering.

It’s a brutal cycle. You spend years reviewing the worst parts of the internet—beheadings, abuse, the stuff that gives you PTSD—just to help build an algorithm that eventually decides you’re too expensive to keep around.

As builders, we love automation. We want to script everything. But when "automation" is used as a PR-friendly word for "firing 1,000 people to save a few cents per hour," it leaves a bad taste. Especially when those people were earning less than $2 an hour to look at things no human should ever see.

Building on Quicksand

There’s a "No gree for anybody" mindset you need when building in the Nigerian ecosystem, or anywhere in Africa for that matter. You have to be paranoid. You can’t just assume the big contract is a sign of success.

I’ve seen it happen in tech hubs from Owerri to Lagos. A startup gets a massive partnership with a bank or a telco. They scale up, hire everyone in sight, move into a fancy office with a beanbag chair and a PS5. Then the "partner" changes their internal strategy, and the startup is left holding a massive bill with zero revenue.

If your business model can be destroyed by one person in a boardroom in California having a bad Tuesday, you’re not building a business; you’re building on quicksand.

The lawsuit these moderators are bringing—seeking $1.6 billion—is a massive "hold my beer" moment for the industry. It’s a signal that the era of using Africa as a cheap, disposable testing ground for AI training might be hitting a wall.

I’m tired of seeing our talent treated like a variable cost that can be deleted with a git branch -D. We need to build products that we own, on platforms we control, serving markets that actually value the human effort behind the screen. If we don't, we're just waiting for the next "redundancy notice" to hit our inboxes.