Mastercard pilots AI-agent shopping and checkout with Swiggy Instamart, Zepto and Tira

Mastercard is testing LLM-led shopping and checkout flows with Indian retailers as brands prepare for AI agents to reshape product discovery. The shift raises the importance of machine-readable catalogues, payment integrations and measuring recommendation share in chatbot results.

— Source published Wed, 19 Aug, 2026, 09:06 IST · First seen Wed, 19 Aug, 2026, 10:35 IST · Source ET BrandEquity

What happened

Mastercard is piloting LLM-based shopping and checkout with Indian retailers including Swiggy Instamart, Zepto and Tira, while extending payment partnerships.

Key facts

  • Oreo appeared in 1 in 10 AI chatbot cookie recommendations
  • AI agents may influence most purchase decisions in 5-10 years, according to Reliance Ajio
  • Bots initiate 57.4% of search requests versus 42.6% from humans

Why this matters

Target partnerships or acquisitions in product-information management, conversational commerce and payment orchestration to secure distribution in AI-agent shopping ecosystems.

What to watch

  • Expansion of Mastercard's India pilot from assisted shopping to autonomous or pre-authorized checkout.
  • Launches of retailer APIs exposing live inventory, substitutions, delivery capacity and loyalty benefits to third-party agents.
  • Network rules defining liability for agent-initiated fraud, mistaken purchases, refunds and chargebacks.
  • Evidence that AI-originated baskets have higher conversion, larger basket sizes, lower acquisition costs or different category mixes than app-originated orders.
  • Retail-media products that let brands pay for sponsored placement or optimized recommendation visibility within agent responses.
  • Consumer adoption of persistent shopping preferences, delegated spending caps and payment credential vaults.
  • Merchant complaints about loss of customer data, opaque recommendation ranking or rising intermediary fees.
  • Regulatory guidance in India on consent, data sharing, AI-generated recommendations and automated payment authorization.
  • Build an agent-ready product data layer with normalized titles, attributes, variants, ingredients, compatibility data, images, reviews, prices, promotions and real-time inventory.
  • Create APIs for search, basket construction, substitution preferences, delivery slots, loyalty redemption, order tracking, returns and customer support.
  • Instrument a new funnel: assistant impressions, recommendation share, citation rate, agent-to-cart rate, agent checkout conversion, substitution rate, cancellation rate and repeat purchase.
  • Prioritize high-frequency, low-consideration categories for pilots, including grocery replenishment, household essentials, personal care, beauty replenishment and gifting.
  • Set agent commerce policies for customer authorization, payment tokenization, spending limits, fraud controls, order confirmation, dispute handling and accountability for recommendation errors.
  • Treat recommendation share in leading LLMs as a commercial channel alongside search ranking, marketplace visibility and retail-media share of voice.
  • Negotiate with payment networks and agent providers for access to transaction data, attribution rules, customer ownership and acceptable merchant fees before volume scales.

Also reported by