EXL buys iMerit to boost enterprise AI

EXLService Holdings Inc. completed its acquisition of iMerit Technology this month. The deal merges iMerit's AI model training and data annotation expertise with EXL's enterprise services portfolio.

Founded in 1999, EXL provides services to industries including insurance, healthcare, banking, retail, and energy. The New York-based company employs about 68,000 people worldwide. iMerit, founded in 2012 and based in San Jose, Calif., specializes in data annotation for robotics, autonomous mobility, and healthcare AI. Its Ango Hub platform and subject-matter experts help create validated training data for complex models.

Acquisition aims for trustworthy enterprise AI

Radha Ramaswami Basu, iMerit's founder and CEO, has joined EXL as an executive vice president. She and EXL Chairman and CEO Rohit Kapoor discussed the rationale with The Robot Report. Basu stated that while AI models are becoming more capable, issues like hallucination and trust remain critical for enterprise deployment. She argued that expert human feedback is essential to challenge models and evaluate their reliability in specific business contexts.

Kapoor explained the acquisition connects previously separate parts of the AI lifecycle. iMerit contributes expert-led model training, evaluation, and its Ango Hub platform. EXL brings deep industry context and experience in integrating technology into business operations. The combined entity aims to offer a continuous path from data preparation to model operationalization within enterprise workflows.

Focus shifts to business value and safety

Kapoor noted that many organizations struggle to achieve consistent business value from AI experiments. He said the challenge involves training, evaluating, adapting, and governing AI systems for specific business contexts. According to Kapoor, this makes capabilities like model evaluation and domain-specific data foundational, not peripheral.

Basu applied this to physical AI, such as robotics and autonomous vehicles. She said success in these fields depends more on data quality than model scale. Physical AI must interpret noisy inputs and act safely in unpredictable environments. Human experts are critical for creating high-quality training data, building realistic scenarios, and identifying edge cases.

For safety and compliance, Basu argued these cannot be a final check. They must be designed into the data, training, and evaluation process from the start. In regulated industries, organizations must demonstrate an AI system was tested against scenarios where an error could cause irreversible harm. EXL's trace analysis capabilities aim to help organizations understand the data and model behavior behind specific decisions.

Competitive advantage seen in combined layers

Kapoor outlined his view of the evolving AI market. He said competitive advantage in enterprise AI will not come from any single technology layer. It will come from combining proprietary data, domain context, model evaluation, AI engineering, and governance. Over the next three to five years, he believes proprietary enterprise data will become a highly valuable competitive asset.

Companies will increasingly want specialized models trained on their own data, Kapoor predicted. They will need continuous evaluation and reinforcement learning to keep models accurate as conditions change. Domain expertise will matter more, not less, especially in complex, regulated workflows like underwriting or fraud detection.

The companies that succeed, according to Kapoor, will be those that combine data, context, AI, evaluation, and execution into one operating model. The iMerit acquisition is presented as a step to strengthen EXL's position in this landscape.