EXL just paid $310M to own the last mile of enterprise AI

EXL just bought iMerit for $310M.

Not for the code. For the humans who train the models.

iMerit runs a global network of physicians, scientists, and engineers who do reinforcement learning from human feedback. The same process that makes foundation models actually work in production.

EXL now has direct relationships with the companies building GPT-5, Claude 4, and Gemini 3. They see how models are trained before anyone else does.

This is the enterprise AI play nobody is talking about. Everyone is fighting over inference. EXL just bought the training layer.

Your CISO approved the model. Your procurement team signed the contract. But nobody asked who trained it, what data they used, and whether the "alignment" was done by a real expert or a crowd-sourced crowd-sourcer.

EXL's 67,000 employees just got a new capability: the ability to fine-tune domain-specific models for regulated industries using verified subject matter experts. Healthcare. Insurance. Banking. Capital markets.

This acquisition quietly solves the hallucination problem for enterprises that cannot afford to be wrong.

Audit your model training pipeline. If you don't know who did the RLHF on your production models, you have a liability problem that no vendor contract can fix.