RESEARCHYou Can't Buy Taste. We Tested It.

Scale AI vs iMerit: crowd economics against payroll economics

The difference between these two companies shows up in their org charts before it shows up in anything they sell. Scale built a data engine on a flexible contributor pool and does not publish how many people are in it. iMerit employs its annotators full time, with benefits, and says so on the homepage because it is the product. Those two choices propagate into consistency, into cost structure, into which verticals each one is credible in, and into what happens when your project needs the same reviewer to see the same edge case for the fourteenth time. This comparison works through where each model wins.

Key takeaways

  • iMerit employs its workforce full time rather than contracting it. Scale publishes no contributor headcount, geography, or pay rates at all.
  • iMerit holds the broadest published compliance set in this category: SOC 2 Type 2, ISO 27001, ISO 9001, HIPAA, GDPR, and TISAX. Scale holds SOC 2 Type II, ISO 27001, FedRAMP High, and DoD IL4, with no HIPAA published.
  • For regulated clinical, geospatial, or automotive work, iMerit's certification spread and in-house subject-matter experts are the better fit.
  • For defense and federal work, Scale's authorizations end the discussion.
  • iMerit was acquired by EXL Service in August 2026. Ask what that changes about pricing, roadmap, and your delivery team.
  • Neither publishes IP indemnification terms.

About iMerit

Founded in 2012, headquartered in San Jose with delivery centres in New Orleans, Kolkata, and Bengaluru, and operations in Bhutan. Its headline differentiator is employment model: iMerit states that its entire workforce across India, Bhutan, and the United States consists of full-time staffers with healthcare and benefits, explicitly not a crowd or gig pool. It reports more than 10,000 active resources across 60-plus countries and an iMerit Scholars network of 25,000-plus domain experts including physicians, scientists, engineers, and linguists.

The service covers multimodal annotation through its Ango Hub platform spanning image, video, LiDAR, DICOM, text, PDF, and audio; RLHF and alignment for language and vision models; foundation model evaluation and benchmarking; a Deep Reasoning Lab for chain-of-thought and sequential reasoning tasks; and red-teaming. It claims output accuracy above 98 percent and is available through AWS Marketplace as both data labeling and human review services.

In June 2026 EXL Service agreed to acquire iMerit for 170 million dollars upfront plus up to 140 million in earnouts over two years, completing the deal on 3 August 2026. Founder Radha Basu joined EXL as an executive vice president and committee member, and iMerit retains its brand as a distinct unit inside a roughly 68,000-employee organization.

About Scale AI

Founded around 2016 by Alexandr Wang and Lucy Guo, based in San Francisco. Meta invested roughly 14.3 billion dollars for a 49 percent stake in June 2025 at a valuation above 29 billion. Francis deSouza became chief executive in August 2026.

Scale sells a Data Engine for annotation and curation across text, image, video, and 3D and LiDAR; a generative AI portfolio covering RLHF, prompt-response generation, and red-teaming; public model leaderboards; and Donovan, a decision-support platform for defense and enterprise.

Its government business is the most verifiable part of its record. Army contract W911QX20C0051 shows 105.5 million dollars total obligated. An Army indefinite-delivery contract signed in May 2026 carries a ceiling near 99.5 million. A February 2026 Air Force award of roughly 5.1 million integrates Donovan into an airborne command platform. The widely cited 500 million dollar Department of Defense agreement is a ceiling, not obligated spend.

What buyers report

Both review bases are small, and iMerit's is the more usable of the two despite being thin.

iMerit holds 4.8 out of 5 on G2 across 12 reviews, with reviewers citing the Ango Hub interface, AI-assisted labeling tools that speed workflows, multimodal cross-format synchronization, and scaling to millions of images. Criticisms are minor interface gripes and an initial learning curve. Capterra shows 5.0 from a single review and PeerSpot has none, so confidence is limited by volume rather than by sentiment.

Scale's GenAI Platform holds 4.5 out of 5 on Gartner Peer Insights from two reviews, citing model quality and tooling, with pricing structure flagged as a concern.

Twelve reviews is thin. Two is an anecdote. Run a pilot.

The comparison

iMeritScale AI
Founded2012Around 2016
WorkforceFull-time employees with benefits, stated explicitlyNot published
Scale of network10,000-plus active resources; 25,000-plus ScholarsNot published
Delivery locationsSan Jose, New Orleans, Kolkata, Bengaluru, BhutanNot published
ModalitiesImage, video, LiDAR, DICOM, text, PDF, audioText, image, video, 3D and LiDAR, multimodal
ComplianceSOC 2 Type 2, ISO 27001, ISO 9001, HIPAA, GDPR, TISAXSOC 2 Type II, ISO 27001, FedRAMP High, DoD IL4. No HIPAA found
GovernmentNone publishedVerifiable federal contract history
Open dataMinimal; a medical imaging dataset and a few models36 datasets and 5 models on Hugging Face
Governance documentsPublished AI Ethics Policy and Environmental and Social PolicyNot published
OwnershipAcquired by EXL Service, August 2026Meta holds 49 percent since June 2025
Reviews4.8 out of 5, 12 reviews4.5 out of 5, two reviews

The employment model is the whole argument

Most vendors in this category describe their workforce in adjectives. iMerit describes it in employment terms, and that is a structural claim a buyer can reason about rather than take on trust.

Full-time employment buys you continuity. The same annotators see the same task types across months, which is what produces consistency on long-running programs, and it is why in-house models tend to outperform flexible pools on work where the edge cases are the point: clinical imaging, geospatial boundaries, autonomous vehicle corner cases. It also means the people doing your work can be retrained rather than replaced, which matters when your taxonomy changes mid-program, as taxonomies do.

Scale publishes nothing comparable. No contributor headcount, no delivery geography, no pay rates. Third-party analysis of its contributor platform reports pay compressing from roughly 40 to 50 dollars an hour in 2024 to roughly 15 to 20 by late 2025, but that is aggregated secondary sentiment rather than company disclosure, and it should be weighted accordingly. What a buyer can say with confidence is narrower and still useful: on the question of who does your work and under what terms, one of these companies answers and the other does not.

Compliance breadth against compliance depth

iMerit publishes the broadest certification set in this category on a single page: SOC 2 Type 2, ISO 27001, ISO 9001, HIPAA, GDPR, and TISAX, the automotive security standard. It also publishes an AI Ethics Policy and an Environmental and Social Policy, both dated January 2025, which is more governance documentation than almost anyone else here offers.

Scale's set is narrower but goes somewhere iMerit's does not. SOC 2 Type II and ISO 27001 overlap, and then Scale adds DoD IL4 and FedRAMP High, usable across NIPR, SIPR, and JWICS networks. It publishes no HIPAA attestation, which is a genuine gap given Mayo Clinic appears as a named customer.

The practical reading: if your work is clinical, automotive, or otherwise regulated outside the federal space, iMerit's spread covers more of what your security review will ask for. If your work touches defense, Scale is the only one of the two that can be used at all.

The acquisition question

EXL completed its acquisition of iMerit on 3 August 2026, for 170 million dollars upfront with up to 140 million in earnouts tied to the following two years. iMerit retains its brand as a distinct unit and its founder joined EXL's executive committee, which is the structure most likely to preserve continuity.

It is still an integration, and integrations move pricing, roadmaps, and delivery teams. The earnout structure is worth understanding specifically, because incentives over a two-year horizon shape which customers get attention. Ask who owns your account after integration, whether your delivery team stays intact, and what the contractual position is if terms change mid-engagement.

Scale has its own version of this question in a different form. Meta's 49 percent stake is permanent until it changes, and in the weeks after the June 2025 investment Google, OpenAI, and xAI were reported to have pulled back or halted projects. If Meta competes with you, that is the first thing to weigh.

Which to choose

Choose iMerit when the work is clinical, geospatial, or automotive; when consistency across a long program matters more than peak throughput; when your security review wants HIPAA and TISAX alongside the usual two; or when you want to know who is doing the work and under what employment terms. Ask about post-acquisition continuity.

Choose Scale when you are deploying into defense or government, when you need breadth across modalities in a single vendor, or when you want published open benchmarks to evaluate before committing. Weigh the Meta stake against your own competitive position.

Why HumanSignal is the better choice when the data has to be created

Both companies here are annotation businesses at their core. iMerit does it with employed subject matter experts and the broadest compliance set in the category; Scale does it at breadth with government authorizations nobody else holds. Both start from data you already have. Neither is built for the case where the examples your model needs have never been recorded, which is now the binding constraint in robotics, embodied AI, and any domain where expertise lives in physical practice.

HumanSignal Services starts there. We write the collection protocol, recruit for the physical or professional characteristics it requires, run the capture in our studios, on site at your facility, or in the field with synchronized motion, vision, audio, and sensor streams, and deliver a dataset verified against the spec we agreed. Accountability for the dataset sits with us rather than with a labeling queue.

The quality mechanism is inspectable. Review workflows run in Label Studio Enterprise, the commercial edition of the open source labeling tool we build, so inter-annotator agreement, reviewer decisions, and rework rates are reportable to you while the work is running rather than summarized afterward. We publish contributor pay ranges, which makes our sourcing claims checkable. Our customer agreement gives you exclusive ownership of customer data and of insights produced specifically for you. We hold SOC 2 Type 2 and HIPAA compliance, documented on our security page.

The capture operation came in through the acquisition of Erud AI in November 2025. If your constraint is producing data rather than labeling it, physical AI data services sets out how a program gets scoped.

Does iMerit's full-time employment model cost more

iMerit does not publish pricing, so a direct comparison is not possible from public sources. In principle an employed workforce carries higher fixed cost and lower marginal cost than a flexible pool, which tends to favor it on long, steady programs and disfavor it on short bursts. Ask both vendors to quote the same scope and compare total program cost including your own management overhead.

What changes for iMerit customers after the EXL acquisition

Publicly, little: iMerit keeps its brand as a distinct unit and its founder joined EXL's executive committee. What is not public is how the two-year earnout structure shapes account priorities. Ask who owns your account post-integration and whether your delivery team is contractually stable.

Which is better for medical imaging

iMerit, on published evidence. It supports DICOM natively, holds HIPAA compliance, publishes medical imaging case-study work, and staffs clinically trained reviewers. Scale publishes no HIPAA attestation despite naming Mayo Clinic as a customer, which is the gap to raise if you are considering it for clinical data.

Can either collect physical or robotics data

Both annotate it, including LiDAR and 3D point clouds. Neither publishes a designed collection capability with its own capture facilities. If you need the data produced rather than labeled, look at collection specialists.

Related Content