Agentic AI is shifting B2B commerce from product search to intelligent decision-making, according to Nikhil Prabhakar, CIO at IndiaMART InterMESH, in an interview published by CIO.com. At IndiaMART, a platform hosting 129 million live product listings, the technology is already handling buyer-supplier matching, voice interactions, and lead qualification at scale. The company is now building agents that can reply to sales inquiries on behalf of sellers, using past conversation patterns to determine tone and approach.
IndiaMART's voice AI agent, called VANI, processes more than 100,000 buyer calls across multiple languages and accents. The platform fields over 29 million unique business inquiries annually, creating a major follow-up challenge for suppliers juggling sales, production, and collections. AI-suggested replies and one-tap templates have increased buyer responsiveness by three times. The company is developing automated response systems that would allow sellers to control how much an agent can send without human review, drawing on previous conversations in Lead Manager to replicate how each supplier quotes, clarifies, and follows up with prospects.
According to Prabhakar, the real promise of agentic AI in B2B commerce isn't just helping buyers find products but understanding their needs and connecting them with the right suppliers. The platform now assesses verification status, response rates, and ratings before a buyer contacts a seller, shifting value toward verified business data as agents take on more tasks. Prabhakar emphasizes that faster activity isn't better selling—fewer missed inquiries and more qualified conversations are what drive productivity. Two barriers remain before SMEs can fully adopt agentic tools: incomplete catalogues that limit what agents can accomplish, and confidence that an agent understands their products and operates within acceptable boundaries.
The company is embedding AI capabilities directly into existing workflows rather than requiring small businesses to learn new models or prompts. AI removes backgrounds, crops images, and extracts attributes to simplify catalogue creation, while bots convert naturally stated needs into structured buying leads through voice and vernacular interfaces. AI-prioritized leads and seller analytics on competition and pricing deliver enterprise-grade intelligence previously available only to large firms. As agents gain more autonomy, Prabhakar argues that verification must become more rigorous, not less—business attributes should be checked at source through GST and MCA data, with AI-assisted video KYC adding another layer. Wrong matches typically stem from vague specifications or incomplete data, he notes, and verified facts must be separated from supplier claims and model inferences.
Looking ahead, Prabhakar says CIOs must govern how software exercises delegated authority, treating agents like employees with identities, permissions, oversight, and audit trails. He recommends starting with measurable workflows, building layered fraud controls and GST-based seller verification into every release cycle, and preparing teams to specify what agents should do and verify what they did. The goal is combining greater autonomy with dependable value and clear responsibility, he concludes. For platforms managing millions of transactions, the shift means standardized, repeat purchases will move to agents first, while complex procurement will keep human judgment at the center.