How Hexaware Uses Elastic AI Assistant to Cut Engineer Training Time by 75%

Hexaware has staked its competitive positioning on being genuinely AI-first, not just claiming the label. With 80% of its workforce certified in advanced AI skills and $1.3 billion in annual revenue, the India-based IT services firm manages complex infrastructure and applications for enterprise clients around the world. The scale of that commitment made the gaps in its day-to-day operations all the more visible: getting new engineers productive took up to 18 months, and monitoring client environments required constant attention from senior staff.

The core problem was the density of institutional knowledge required to run managed services. Engineers arriving from university had to memorize hundreds of error codes, navigate dozens of dashboards to assess the health of client systems, and rely on lead engineers for guidance on complex incidents. This bottleneck was expensive in time and in senior talent, and it slowed Hexaware’s ability to deploy new hires effectively.

Hexaware placed Elastic at the center of its AI operations roadmap, deploying Elastic Observability for anomaly detection and alert management, and Elastic AI Assistant as a generative AI layer that lets engineers query system data in natural language. The AI Assistant pulls context from live operational data so that when an engineer clicks on an unfamiliar error code, it surfaces causes, context, and recommended next steps—without requiring memorization of technical manuals.

The most striking change came in talent deployment. Through its Mavericks Program, Hexaware now onboards university graduates and has them working on client projects within three months rather than the previous 12 to 18. Reliance on lead engineers dropped by up to 90% for routine queries. At the same time, teams monitoring client environments saw operational efficiency improve by 50%—instead of checking dozens of dashboards, an engineer can ask in plain English how a service performed overnight and receive a structured answer in seconds. In one Elastic Observability deployment, weekly false positive alerts dropped from 1,000 to 244.

Looking ahead, Hexaware is extending these patterns to more of its client work. In a deployment for a European energy company, Elastic’s ELSER semantic retrieval model helped reduce mean time to resolve by 40% by automatically matching incoming incidents to historical cases. A US healthcare provider saw a similar pattern: ELSER-powered query handling improved accuracy to 90% while cutting query times by 40%. Hexaware’s bet is that generative AI embedded in observability infrastructure will progressively reduce the demand for human agents in call center and support contexts—turning experienced engineers into reviewers of AI-generated work rather than first responders.