ExlService Holdings (EXLS ), better known as EXL, has completed its acquisition of AI data specialist iMerit, bringing together two companies focused on different but closely connected parts of the enterprise artificial intelligence stack.
The transaction, originally announced in June, was valued at up to $310 million. It includes $170 million in upfront consideration and up to $140 million in incentives and earnouts tied to milestones over the next two years. iMerit founder and CEO Radha Ramaswami Basu will join EXL as Executive Vice President and Head of iMerit, while also becoming a member of EXL’s executive committee.
Addressing the Gap Between AI Pilots and Production
The acquisition is intended to address one of the most persistent problems in enterprise AI: moving a promising model or proof of concept into a dependable production system.
Developing an AI application involves considerably more than selecting a foundation model. Enterprises must prepare and govern their data, fine-tune models for specific use cases, evaluate outputs, identify failure modes, establish guardrails, and continuously monitor performance after deployment.
These requirements become more demanding in industries such as insurance, healthcare, banking, and capital markets, where inaccurate outputs can create financial, regulatory, or operational consequences.
EXL has positioned itself around the enterprise side of this process, combining data management, analytics, industry expertise, and AI-powered operations. iMerit brings capabilities closer to the model-development layer, including training data creation, model evaluation, reinforcement learning, red teaming, and expert human feedback.
The combination gives EXL a more complete set of capabilities extending from preparing enterprise data to evaluating how models and AI agents behave in real business environments.
What iMerit Adds to EXL’s AI Portfolio
Founded by Radha Basu, iMerit specializes in producing and evaluating the high-context data required to train large language models, vision models, multimodal systems, and other advanced AI applications.
Its generative AI services cover prompt-and-response creation, supervised fine-tuning, reinforcement learning from human feedback, corpus development, retrieval-augmented generation fine-tuning, safety evaluation, red teaming, and AI agent testing.
Much of this work combines software with expert human review. Through its iMerit Scholars network, the company sources domain specialists such as physicians, scientists, engineers, and linguists who can evaluate model behavior in areas where general-purpose annotation may be insufficient.
This expert-in-the-loop approach is particularly relevant when models must understand specialized terminology, follow industry rules, assess complex reasoning, or operate within regulated workflows. iMerit also works across text, images, video, audio, LiDAR, medical imaging, and other data formats, giving EXL additional capabilities for multimodal AI projects.
Ango Hub Brings an Evaluation and Training Layer
A central part of the acquisition is iMerit’s Ango Hub, a platform for managing AI data workflows and interactions between models, annotators, and subject matter experts.
The platform supports configurable workflows for data annotation, prompt-response generation, chain-of-thought reasoning tasks, quality assurance, model alignment, and reinforcement learning. It can also integrate with customers’ existing infrastructure through application programming interfaces, custom schemas, and private deployments.
For AI agents, Ango Hub can be used to inspect tool calls, score agent behavior against structured rubrics, review execution traces, and identify issues such as planning errors, tool misuse, memory drift, or policy violations.
These capabilities reflect the growing need to evaluate not only what an AI system says, but also the actions it takes and the sequence of decisions leading to an outcome.
As enterprises move toward more autonomous AI systems, this kind of evaluation infrastructure may become an important control layer between experimental agents and production deployment.
How iMerit Fits Into EXL’s Existing Technology Stack
EXL has been expanding its proprietary AI portfolio through platforms such as EXLerate.ai and EXLdata.ai.
EXLerate.ai is the company’s agentic AI platform for developing and deploying AI systems within enterprise workflows. The platform includes prebuilt agents and accelerators, along with governance, auditing, cost controls, and support for regulated industries. The company has also incorporated NVIDIA (NVDA ) technologies for model development, agent orchestration, and accelerated computing.
EXLdata.ai focuses on an earlier part of the AI lifecycle: converting fragmented, unstructured, or poorly governed enterprise information into trusted, AI-ready data. Its modules cover areas such as data migration, quality, governance, annotation, and DataOps, while integrating with platforms including Databricks, Snowflake, and Amazon Web Services.
iMerit adds a complementary model training and evaluation layer. In practical terms, EXL can now help clients organize proprietary data, configure AI models and agents, generate expert training signals, evaluate performance, and integrate the resulting systems into business operations.
Radha Basu Joins EXL’s Executive Committee
Basu’s appointment gives iMerit direct representation within EXL’s senior leadership rather than positioning the acquired company solely as a new business unit.
In her new role, Basu will lead iMerit while helping shape EXL’s broader AI strategy. Her background sits at the intersection of technology, workforce development, and human-in-the-loop AI, an area that is becoming more significant as companies look for methods to improve model accuracy without removing human oversight.
EXL Chairman and CEO Rohit Kapoor described the transaction as a shift that will deepen the company’s vertically specialized AI capabilities. Basu, meanwhile, emphasized that enterprise AI success will depend less on which underlying model a company selects and more on how effectively that model can be deployed within real business environments.
Building a More Complete Enterprise AI Platform
The acquisition reflects a broader change in the enterprise AI market. The initial race to experiment with large language models is giving way to more operational questions involving data quality, evaluation, compliance, cost, integration, and measurable performance.
Buying access to a model is relatively straightforward. Ensuring that it works consistently with proprietary data, follows internal policies, handles unusual cases, and remains reliable after deployment is considerably more difficult.
By combining EXL’s enterprise data and operational capabilities with iMerit’s model training, evaluation technology, and domain-expert network, the companies are attempting to cover more of that lifecycle within a single organization.
The acquisition does not eliminate the technical and organizational challenges associated with deploying AI at scale. Enterprises will still need clearly defined use cases, reliable data, governance processes, security controls, and methods for measuring business outcomes. However, it gives EXL a broader foundation for helping clients manage those requirements as AI projects move beyond experimentation and into production.
