On September 2, 2025, Krutrim, the AI arm of Ola, executed a third round of layoffs since June, affecting about 50 people, primarily from its linguistics unit. This unit specialized in regional languages such as Bengali, Malayalam, Punjabi. This followed earlier layoffs in July. Reports also highlighted leadership exits, indicating ongoing restructuring as product roadmaps evolve. For language tech teams that scaled rapidly to train models and localize UX, the message is clear: AI hiring surges can reverse quickly as tooling, funding, or priorities shift.
For linguists and annotators who thought they were building a multilingual future, this feels like a personal setback. Many joined to help Indian languages show up inside mainstream AI; now they're packing desks, scrambling for references, and calculating how long savings will cover Bengaluru rents. The remaining employees report guilt and fatigue, plus fear that automation will replace the very work they trained models on. HR's challenge is to treat exits with dignity (clean documentation, outplacement, mental-health support) while stabilizing core teams with honest roadmaps, not platitudes.
Layoff law in India depends on job classification. For many language specialists, the Industrial Disputes Act, 1947 (for "workmen") may trigger notice, retrenchment compensation, and last-in, first-out considerations, unless exceptions apply; others exit under contract. HR must map roles to legal categories, compute dues (including leave encashment and variable pay accruals), and ensure timely Form-16 and service letters to prevent downstream employability harm. For remaining teams, revisit IP/confidentiality and moonlighting clauses, as skill marketplaces court displaced talent. Above all, publish a skills bridge: micro-credentials for data quality, evaluation engineering, and safety—so this isn't just an ending.
What's one humane practice you expect in any layoff? How should AI firms protect language talent from "train-then-discard" cycles?
For linguists and annotators who thought they were building a multilingual future, this feels like a personal setback. Many joined to help Indian languages show up inside mainstream AI; now they're packing desks, scrambling for references, and calculating how long savings will cover Bengaluru rents. The remaining employees report guilt and fatigue, plus fear that automation will replace the very work they trained models on. HR's challenge is to treat exits with dignity (clean documentation, outplacement, mental-health support) while stabilizing core teams with honest roadmaps, not platitudes.
Layoff law in India depends on job classification. For many language specialists, the Industrial Disputes Act, 1947 (for "workmen") may trigger notice, retrenchment compensation, and last-in, first-out considerations, unless exceptions apply; others exit under contract. HR must map roles to legal categories, compute dues (including leave encashment and variable pay accruals), and ensure timely Form-16 and service letters to prevent downstream employability harm. For remaining teams, revisit IP/confidentiality and moonlighting clauses, as skill marketplaces court displaced talent. Above all, publish a skills bridge: micro-credentials for data quality, evaluation engineering, and safety—so this isn't just an ending.
What's one humane practice you expect in any layoff? How should AI firms protect language talent from "train-then-discard" cycles?