Traditional managed-workforce providers: Appen, iMerit, Sama, TELUS International, CloudFactory

The BPO-style annotation layer: what they are and why they still matter

Here's the thing, when everyone's talking about Scale's government contracts or Surge AI's expert workforce, it's easy to forget that the traditional managed-workforce providers, Appen, TELUS International, iMerit, Sama, CloudFactory, still handle massive volumes of annotation work. These companies came out of the BPO (business process outsourcing) world, built tens of thousands of human workers before anyone was fine-tuning GPT-4, and they're still winning deals that need global reach, multilingual coverage, or domain specialization.

What sets them apart is the full-stack managed service model, they own recruitment, training, quality control, and workforce management end-to-end. You don't bring your own tooling or manage contractors yourself; you send them a dataset spec, and they send back labeled data. Appen has 30,000+ multilingual workers across 170+ countries. TELUS International (which acquired Lionbridge) has 79,000 employees with deep localization expertise. iMerit focuses on domain experts in medical imaging and geospatial data. Sama positions as a B-Corp with ethical sourcing from East Africa and guarantees 95-99.5% accuracy. CloudFactory runs a 7,000-person managed workforce processing 50,000 daily labels for retail and e-commerce clients.

From what I've seen in PM interviews, the mistake is dismissing these players as "obsolete" because they're not the hot new expert-tier platforms. In reality, they complement rather than compete with Surge or Mercor. If you need 500+ languages covered for a global product launch, or you need to label 50,000 hours of driving video in three months for an AV dataset, or you need medical imaging annotations from radiologists in India, these are the providers you call. They handle scale, specialization, and operational complexity that newer platforms deliberately avoid.

Check your understanding

A director-level PM at an autonomous vehicle company needs to label 100,000 hours of driving footage across 15 international cities in 4 months. The dataset includes pedestrian tracking, traffic sign recognition, and lane markings. Which provider type is the best strategic fit?

Where each player wins: workforce composition and strategic wedges

The traditional providers aren't interchangeable, each has a distinct workforce composition and go-to-market wedge that determines where they win. Appen's superpower is multilingual crowd scale, 170+ countries, 500+ languages, 30,000+ contractors. If you're launching a voice assistant in 80 languages or need sentiment analysis across global markets, Appen's the default. TELUS International (Lionbridge) comes from the localization and translation world, so they dominate when annotation overlaps with content localization, think labeling UI elements for international rollouts or moderating multilingual content for trust and safety.

iMerit carved out domain-expert niches, medical imaging (radiologists), geospatial data (satellite imagery analysts), and autonomous vehicles. They're not trying to be the cheapest; they're selling specialized accuracy for regulated or high-stakes domains. Sama differentiated on ethical sourcing and impact, they're a B-Corp, they hire from underserved communities in East Africa, and they offer a 99.5% accuracy SLA that appeals to enterprises with compliance requirements. CloudFactory runs a managed workforce of 7,000 people focused on retail, e-commerce, and customer support use cases, processing product catalogs, moderating reviews, labeling inventory images.

What's useful for you as a PM is understanding that these wedges determine build-vs-buy trade-offs. If your AI product needs multilingual coverage at scale, building internal tooling won't solve the workforce recruitment problem, Appen already has those contractors. If you need radiologists to label chest X-rays, iMerit's domain network is the asset, not their annotation UI. The platform is almost secondary; the workforce composition is the competitive moat.

The workforce is the moat, not the tooling

Check your understanding

Match each traditional annotation provider to the workforce wedge or strategic positioning that defines where they win deals.

How they fit into the modern annotation stack, and where they're losing ground

Let's get real, traditional BPO-style providers are not obsolete, but they're also not the default choice anymore for cutting-edge AI labs. The biggest shift is that expert-tier platforms like Surge AI and Mercor captured the high-value RLHF and alignment work that powers frontier models. Meta's own researchers reportedly view Scale data (which itself operates a managed workforce model) as lower quality than Surge or Mercor, and that perception extends to traditional providers. When you're fine-tuning GPT-4 or building Claude, you want expert contractors at $50-$100+/hr who can evaluate nuanced reasoning, not crowd workers paid $5-$15/hr clicking bounding boxes.

The second challenge is that self-serve platforms like Labelbox and SuperAnnotate unbundled the tooling from the workforce. If you already have an internal annotation team or you want to hire contractors yourself, you don't need Appen's end-to-end managed service, you just need software. That's why SuperAnnotate's drag-and-drop UI is ranked #1 on G2 for computer vision: it gives you control over iteration speed and quality without paying for workforce management overhead.

But here's where traditional providers still win: operational complexity at global scale. If you need to label datasets in 80 countries with local compliance requirements, or you need 24/7 moderation shifts across time zones, or you need to ramp 500 annotators in two weeks for a product launch, that's not a tooling problem, it's a workforce logistics problem. Appen, TELUS, iMerit, and CloudFactory have spent decades solving that. They're less relevant for cutting-edge model training, but they're still the backbone for multilingual products, global trust and safety, and domain-specific enterprise datasets.

Check your understanding

Key takeaways

• Traditional managed-workforce providers (Appen, iMerit, Sama, TELUS International, CloudFactory) handle massive annotation volumes through full-stack BPO models, recruitment, training, QC, and delivery.
• Each provider has a distinct workforce wedge: Appen dominates multilingual scale (170+ countries, 500+ languages), iMerit sells domain experts (medical, geospatial, AV), Sama differentiates on ethical sourcing and compliance SLAs, TELUS International owns localization overlap.
• The workforce composition is the competitive moat, not the annotation tooling, building a 7,000 to 79,000 person workforce internally takes years and massive operational overhead.
• These providers lost ground on cutting-edge RLHF and expert-tier work to Surge AI and Mercor, and they face unbundling pressure from self-serve platforms like Labelbox and SuperAnnotate.
• They still win deals requiring operational complexity at global scale, multilingual products, 24/7 trust and safety moderation, rapid workforce ramp-up, and domain-specific enterprise datasets where tooling alone isn't enough.

Your product check-in

Apply “Traditional managed-workforce providers: Appen, iMerit, Sama, TELUS International, CloudFactory” to a product or workflow you know. What would you try, what could go wrong, and what evidence would help you decide?