Enterprise AI is clearly entering a new phase. After two years dominated by proofs of concept and generative AI pilots, the challenge for technology leaders is no longer whether AI works, but where it creates measurable business value and how it can be embedded into core enterprise workflows without driving unsustainable costs.
For digital-native companies, where customer interactions, engineering velocity and operational efficiency directly influence growth, AI is increasingly becoming part of the operating model itself.
Speaking to Express Computer, Pramod Adiddam, Chief Technology Officer at Myntra, said the company’s AI investments are guided less by technology adoption than by business outcomes. “We don’t start with AI. We start with the business problem. The hypothesis is always around scale, quality and efficiency,” he says.
That philosophy has resulted in an AI strategy spanning customer engagement, seller operations, enterprise workflows and software engineering, with every deployment expected to demonstrate measurable impact before wider rollout.
Moving from features to AI-driven customer journeys
Rather than deploying standalone AI capabilities, Myntra has organised its customer strategy around AI-orchestrated shopping experiences.
One example is Maya, the company’s conversational shopping assistant, which is designed to assist customers throughout product discovery rather than functioning as a conventional chatbot. According to the company, around 40 percent of Maya users are already highly engaged Myntra customers, while 25 percent of those engaged users return for repeat interactions, indicating continued usage beyond initial experimentation.
Personalisation has also expanded beyond recommendations into search. Myntra says nearly 90 percent of its monthly active users now experience personalised search, where AI combines customer intent, browsing behaviour and contextual signals instead of relying solely on keyword matching.
The same approach extends to merchandising. Through AI-powered “Algo Stores”, approximately 80 percent of users now engage with algorithmically curated storefronts, while nearly 60 percent of homepage impressions are personalised, reflecting the growing role of AI in content ranking and product discovery.
Unlike many early AI deployments, these capabilities are subjected to continuous A/B testing before wider rollout. “For any customer feature, whether it is agentic or not, we do A/B tests. We measure whether the hypothesis is actually translating into customer impact. Depending on the results, we either iterate, pivot or sometimes pause before coming back later,” says Adiddam.
The emphasis on experimentation reflects an increasingly common enterprise challenge. As AI deployments scale, the ability to measure business outcomes is becoming as important as the models themselves.
Agentic AI moves into enterprise operations
Perhaps the more consequential shift is occurring away from customer-facing experiences and into enterprise operations. Myntra is deploying AI agents across multiple operational functions, including customer service, merchandising, fulfilment and business intelligence.
Within customer support, agentic voicebots and chatbots are handling routine customer interactions while automating workflows such as returns approvals. In merchandising, AI generates product images and videos, performs catalogue quality checks and identifies corrections. Fulfilment teams use AI for dynamic capacity planning across logistics networks, replacing manual planning with predictive optimisation.
Internally, the company has also developed BIRA, an AI-powered business intelligence assistant that enables employees to perform analysis and root-cause investigations through natural language queries.
Taken together, these deployments point to a broader enterprise trend. Rather than using AI merely as an assistant, organisations are increasingly embedding autonomous decision support into operational workflows where efficiency gains are more measurable.
AI becomes part of software engineering
If customer AI improves engagement, engineering AI determines how quickly those capabilities reach production. Myntra is extending generative AI beyond code generation into the broader product development lifecycle, including technical documentation, implementation planning, pull request analysis and code reviews.
According to Adiddam, the approach has reduced product development timelines by nearly 40 percent, shortening typical delivery cycles from around 10 days to approximately six days.
“We have reduced our product development time by roughly 40 percent. What earlier took around ten days now takes close to six days. That ultimately means shipping more features faster for customers and sellers,” he says.
The company’s AI strategy therefore resembles an engineering acceleration platform rather than a collection of coding assistants.
AI economics become the next competitive advantage
While AI adoption continues to accelerate, enterprises are increasingly confronting another challenge, the challenge of inference costs.
Rather than sending entire code repositories to large language models, Myntra uses contextual retrieval techniques that expose only the relevant portions of code required for a particular task, reducing token consumption without compromising output quality.
“We are very financially responsible. Our technical architecture has been well thought through from the beginning. We optimise token usage by providing only the relevant code context instead of the entire codebase,” says Adiddam.
The focus on cost discipline reflects a broader reality facing enterprise AI. As AI usage expands across organisations, sustainable architecture may prove as important as model performance.
From AI adoption to AI operating models
Perhaps the most notable aspect of Myntra’s strategy is that AI is no longer treated as a collection of isolated initiatives. Instead, the company has organised its roadmap around four interconnected pillars of AI-orchestrated customer journeys, technology-enabled seller management, AI-powered business operations and an AI-native product development lifecycle.
That evolution mirrors a wider shift taking place across digital enterprises. Competitive advantage is moving beyond deploying individual AI applications to redesigning workflows, engineering practices and decision-making around AI.
For enterprise CIOs, the lesson is not that every AI initiative will succeed. Rather, long-term value is likely to come from disciplined experimentation, measurable business outcomes, cost-aware architecture and the ability to integrate AI into core operating processes instead of treating it as a standalone technology project.