AI agent training for operations managers in e-commerce in India

For an operations manager at an Indian e-commerce business, training for AI agents starts with the limits: what an agent may refund, cancel or promise, and who reviews what it holds. Then come reading the workspace, handling incidents and applying the DPDP Act to customer data. Agents triage and draft at scale; refund policy, seller disputes and escalations stay with the manager.

What does an e-commerce operations manager in India deal with?

Cash on delivery still drives much of the work. Customers refuse parcels at the door, orders return to origin, and couriers settle collected cash on their own cycle. The operations manager tracks every failed delivery and every rupee owed, and decides how to cut the rate of refusals without turning away genuine buyers.

Prepaid orders bring their own exceptions: instant payments that succeed for the customer but fail to reach the merchant, refunds that must go back to the original method, and card payments that the business may not store in full. Marketplace sellers add another layer, with listing disputes, late dispatches and quality complaints.

Customers write in many languages, and expectations about speed of response are high. Festive sales multiply volume for a few weeks, and the operations manager has to staff for them.

Where do agents help, and at what depth?

• Delivery exception triage (depth 1). The agent reads courier updates and groups failed deliveries by reason: customer unreachable, address incomplete, refused at door.
• Customer messages in the customer's language (depth 2). The agent drafts replies confirming a refund, explaining a delay or asking for a corrected address. A person sends, until drafts have proved reliable.
• Prepaid refunds for returned items (depth 3). Where the item is back and checked, the agent issues the refund to the original payment method within the business's limits.
• Seller penalty or payout adjustments (depth 4). Changes to what a seller is paid are held for a named person.
• Reversing a settled payment. Withheld at the platform floor for every agent at every depth.

The business's limits are one of the five bounds every action must meet. The lowest bound wins, and the verdict names it, so the operations manager can see whether a refund was held by their own limit or by the agent's certificate.

What must operations managers learn?

First, the mechanics of a governed agent: certified depth and the depth each action needs, the four verdicts, the minimum chain and provenance. A courier's status update is external; the business's own order record is sealed. When they disagree, the agent should hold, and the manager should know why.

Second, how to set and adjust limits. A refund limit too high means losses when something goes wrong; too low means the queue fills during festive sales. Third, how to read the workspace: held volumes, reasons, certificate status, usage and cost. Fourth, what to do when the agent's certificate lapses after 90 days or a run that clears no depth drops it to depth 1 the same day.

The manager also needs to know what never to delegate, set out in what operations managers should never let an AI agent do. Other roles that need training, from customer service to seller support, are in which roles in e-commerce need AI agent training.

What stays with the operations manager?

Refund and cancellation policy and the limits that enforce it. Decisions about sellers: penalties, delistings and payout holds. Escalations: suspected organised fraud, repeated cash-on-delivery refusals from one area, complaints that reach the grievance officer. And the decision to pause an agent.

Agentic commerce makes this sharper. When a customer's own agent places or cancels orders, its instructions are external claims until verified. They cannot authorise a payment, and the operations manager must decide what verification is enough. See what is agentic commerce?.

Which Indian rules shape oversight?

The Digital Personal Data Protection Act 2023 governs the personal data the agent processes: names, addresses, phone numbers, order histories. It requires a clear notice and a lawful basis for processing, use limited to the stated purpose, security safeguards, and a way for people to exercise their rights and raise grievances. The agent's access to customer data should follow those limits. See the DPDP Act and AI agents in India.

The RBI regulates payment systems and payment aggregators, including rules that stop merchants storing full card details and rules on refunds through payment systems. Consumer protection rules for e-commerce require a grievance redressal mechanism, clear return and refund policies and fair dealing with customers.

There is no specific Indian AI law; see how AI agents are regulated in India.

This is an explanation of the rules, not legal advice; confirm how they apply to you with counsel.

How should the first 90 days run?

Days 1 to 30: foundations and observation. The operations manager, team leads, seller support and the grievance officer take Foundations (about 4 hours). The agent triages delivery exceptions at depth 1, and leads compare its groupings with their own.

Days 31 to 60: limits and review. The operations manager and leads take the Operator modules: reading the workspace, usage and billing, when something goes wrong, running the rhythm. Refund limits are written down and agreed with finance. Customer messages move to depth 2.

Days 61 to 90: certification before the season. Exam (40 questions in 60 minutes; 32 correct needed) and two-week practicum. Prepaid refunds move to depth 3 only after the agent is certified on the business's own returns. Finish before festive sales begin.

For where this role fits among others, see AI agent skills by role.