Radha Basu, EVP and head of iMerit (left), and Rohit Kapoor, CEO of EXL (proper). Credit score: Mark Liflander, EXL
The actions of bodily AI, together with robots, are solely pretty much as good because the human experience and information used to construct the AI fashions that prepare them. ExlService Holdings Inc. this month accomplished its acquisition of iMerit Expertise, a pacesetter in AI mannequin coaching, analysis, and reinforcement studying.
Based in 1999, EXL gives providers and options to industries together with insurance coverage, healthcare, banking and capital markets, retail, communications and media, and vitality and infrastructure. The New York-based company has about 68,000 staff worldwide.
Based in 2012, iMerit supplies information annotation for robotics, autonomous mobility, healthcare AI, and different high-tech industries. The San Jose, Calif.-based company stated its subject-matter specialists and proprietary Ango Hub platform allow clients “to collaborate on advanced multimodal information to generate extremely curated and validated coaching artifacts for high-stakes fashions.”
As an example, iMerit has partnered with Carbon Robotics to assist digest hundreds of thousands of plant pictures wanted to create an agricultural mannequin for robotic weeding.
Radha Ramaswami Basu, founder and CEO of iMerit, has joined EXL as govt vp and head of iMerit. She and Rohit Kapoor, chairman and CEO of EXL, replied to the next questions from The Robotic Report concerning the acquisition.
Register now and save in your cross to RoboBusiness 2026iMerit and EXL try for extra reliable fashions
What challenges are AI mannequin builders underestimating as we speak, and why?
Basu: Fashions have gotten extremely succesful daily. With this comes growing potential for hallucination or exceeding the scope. The labs, like clients, are beginning to notice that the ultimate mile and the idea of belief want work with a view to thrive in a crucial enterprise course of.
A mannequin can carry out effectively on a benchmark and nonetheless battle with edge circumstances, unfamiliar situations, or the specialised context of a healthcare or monetary workflow. The bottleneck is entry to professional, domain-specific information and trusted deployment in a enterprise workflow.
That is solved by specialists who can problem the mannequin, expose failure modes, and consider whether or not its reasoning and habits are dependable in a given enterprise setting.
How does EXL’s acquisition of iMerit change how fashions are educated and evaluated?
Kapoor: The acquisition connects components of the AI lifecycle which have usually been managed individually, establishing an end-to-end AI platform for enterprises.
iMerit brings expert-led mannequin coaching, multimodal analysis, the Ango Hub platform, and a world community of area specialists. EXL brings enterprise information, deep trade context, and expertise integrating expertise into enterprise operations.
Collectively, these capabilities create a extra steady path from getting ready proprietary information to fine-tuning a mannequin, evaluating the way it behaves, and operationalizing it inside an enterprise workflow as purchasers transition from pilot to production-scale AI. The objective is to assist organizations bridge the hole between innovation and real-world manufacturing, turning AI potential into measurable enterprise outcomes which are reliable.
Bodily AI nonetheless has to show enterprise worth
Why aren’t different firms taking your strategy?
Kapoor: Enterprise AI is coming into a extra demanding part. Organizations have experimented extensively, however most nonetheless battle to attain constant enterprise worth.
The problem was by no means simply selecting a mannequin; it’s coaching, evaluating, adapting, and governing AI programs to carry out reliably in particular enterprise contexts. That’s the place capabilities like mannequin analysis, reinforcement studying, domain-specific information, purple teaming, and professional human suggestions turn into crucial.
AI economics may even push enterprises towards extra specialised, purpose-built fashions, educated on proprietary information, optimized for their very own workflows, and constantly improved primarily based on efficiency. That makes high-quality information, analysis, and reinforcement studying foundational capabilities, not peripheral providers.
This funding displays a broader market shift: Worth is transferring from constructing AI fashions to creating AI usable, trusted, cost-effective, and outcome-driven.
Human specialists assist assure security, notes iMerit head
Are you able to give examples of how this is applicable to the event of autonomous automobiles and robotics?
Basu: As AI strikes into robotics and autonomous programs on the edge, success relies upon much less on mannequin scale and extra on information high quality. Bodily AI should interpret noisy multimodal inputs, purpose in actual time, and act safely in unpredictable environments to determine the right subsequent motion, not simply the “proper” reply.
The challenges evolve from notion to planning and decision-making. Human specialists play a crucial function in creating high-quality floor fact information, they usually assist construct real looking eventualities, determine uncommon edge circumstances, and consider whether or not a automobile or robotic responded appropriately.
In autonomous driving, AI is educated on multimodal information — imaginative and prescient, lidar, audio — and examined via simulated eventualities like collisions. Specialists consider not simply what the automobile sees, but in addition the way it behaves and explains its selections, making certain it might act safely in real-world edge circumstances. These are advanced, high-context, and high-cognition eventualities.
What are the implications for security and compliance?
Basu: The blocker for AI deployment now could be belief, accuracy, and validation versus mannequin functionality. Thus, security and compliance can’t be handled as a closing test earlier than launch. They should be designed into the info, coaching, analysis, and monitoring course of.
Knowledgeable-driven information and area perception will be remodeled into interplay patterns that assess, torment, and oversee AI fashions in mission-critical functions.
In regulated industries, organizations should be capable of show not solely that an AI system performs effectively on common, however that it was examined in opposition to the conditions the place an error may create irreversible scientific, monetary, or operational hurt. EXL strengthens this course of via hint evaluation, enabling organizations to observe the info and mannequin habits that led to a specific choice and higher perceive the place errors or surprising outcomes originated.
Aggressive benefit comes from tech combos, area experience says EXL CEO
How do you see the AI market evolving, each when it comes to the market and the functions?
Kapoor: Aggressive benefit in enterprise AI gained’t come from any single layer. It would come from bringing the correct layers collectively: proprietary information, area context, mannequin analysis, AI engineering, and governance.
Basis fashions will hold enhancing, however within the enterprise, the query isn’t which mannequin is strongest. It’s which AI system delivers probably the most dependable final result in a selected workflow, beneath actual regulatory and operational constraints. That’s a essentially completely different problem.
Over the following three to 5 years, proprietary enterprise information will turn into one of the crucial precious aggressive property. Corporations will more and more need specialised fashions educated on their very own information, not simply entry to exterior mannequin suppliers. They usually’ll want steady analysis and reinforcement studying to maintain these fashions correct and secure as situations change.
Area experience will matter extra, not much less. In advanced workflows, underwriting, claims, healthcare fee integrity, fraud detection, credit score decisioning, the mannequin wants to know the enterprise, the chance, and the regulatory setting. That’s the place EXL has a differentiated place.
The businesses that win in enterprise AI shall be people who mix information, context, AI, analysis, and execution into one working mannequin. That has been EXL’s technique. The iMerit acquisition strengthens it and positions us to play a extra central function in how next-generation AI is constructed, ruled, and deployed for actual enterprise outcomes.
Editor’s observe: Sudeep George, chief expertise officer at iMerit, not too long ago participated in The Robotic Report‘s webinar on “Physical AI and Robotics.”
The put up EXL acquires bodily AI mannequin developer iMerit appeared first on The Robotic Report.
