Rotavision India

Strategic Guide

Retail & E-commerce

Agent Governance for Indian Commerce

A strategic guide to governing autonomous AI agents across pricing, recommendations, and customer service for 500M+ online shoppers.

Executive Summary

Commerce agents are deciding what you see and what you pay. Pricing agents adjust dynamically across thousands of dark stores. Recommendation agents shape purchasing decisions for hundreds of millions of shoppers in 22 languages. Customer service agents resolve disputes autonomously. Yet most retailers cannot explain a single pricing decision or recommendation. ONDC is creating multi-agent commerce at national scale with no governance playbook. The Consumer Protection Act 2019 demands dark pattern prevention that most platforms haven't operationalised. DPDP Act compliance remains aspirational. This guide provides the agent governance roadmap for Indian commerce.

01The Agent RealityAgents decide. Nobody watches.

02Agents Without OperationsWhat ungoverned agents cause

03Agent Use CasesVernacular, pricing, ONDC agents

04Agent EconomicsCost architecture at scale

05Regulatory FrameworkConsumer protection for agents

06Agent Operations StackRegistry to bounded autonomy

07Production ReadinessFive gates for agents

08The PlatformAgent governance infrastructure

09Agent ImplementationsWhat retailers build

10Getting StartedCommerce Agent Trust Accelerator

rotavision.com February 2026

India's e-commerce is exploding — projected to reach Rs 7 lakh crore by FY25. AI agents now autonomously recommend products, set dynamic prices, manage inventory across thousands of dark stores, and handle customer queries in 22 languages. The shift from static recommendation engines to autonomous commerce agents changes the governance problem entirely.

500M+

Online Shoppers by 2030 — Bain / Flipkart, 2024

60%+

Quick Commerce YoY Growth — Redseer, 2024

Rs 7L Cr

E-commerce Market FY25 — Bain & Company

Four Gaps That Define the Problem

The Recommendation Agent Gap

500 million shoppers by 2030, each served by recommendation agents that decide what products they see, in what order, at what price. No reasoning trace for why a specific product is shown. No fairness monitoring across languages, geographies, or demographics. The agent decides. The customer has no recourse.

The Pricing Agent Gap

Dynamic pricing agents adjust prices in real time across marketplaces, quick commerce, and D2C channels. Surge pricing during demand spikes, personalised discounts based on browsing history, markdowns on perishables — all autonomous. No audit trail. No guardrails against predatory pricing. No consumer explanation.

The Quick Commerce Gap

Quick commerce grew 60%+ year-on-year. Pricing and inventory agents make real-time decisions at 10-minute delivery speed — no time for human review. Dark store coordination, stock transfers, and demand forecasting are fully autonomous. When an agent raises prices during a heatwave, who is accountable?

The ONDC Governance Gap

ONDC is democratising e-commerce through an open protocol where agents from different sellers, logistics providers, and buyer platforms interact. Agent governance across this open network — verifiable identity, cross-platform policy enforcement, reasoning capture — is uncharted territory at national scale.

The Core Problem

This isn't a technology problem. It's an agent operations problem. The agents work. The governance doesn't exist.

Most vendors ship a recommendation engine and call it done. Agentic commerce — where AI reasons, decides, and acts autonomously across pricing, inventory, and customer interactions — needs a fundamentally different governance architecture. The gap between deployed agents and governed agents is where regulatory and consumer risk lives.

Commerce Agents Without Operations

Deploy recommendation agent, validate accuracy on test set
No reasoning trace for why specific products are shown to a customer
Pricing agent changes prices dynamically with no audit trail
Customer service agent hallucinates product features and delivery timelines
No fairness monitoring for recommendations across demographics
ONDC compliance is manual and after-the-fact

Commerce Agents with Rotavision

Every agent registered with autonomy level and consumer impact scope
Reasoning capture for every recommendation and pricing decision
Customer service agents with hallucination detection — never promise what doesn't exist
Fairness monitoring to prevent recommendation bias across demographics
Bounded autonomy for pricing agents — guardrails prevent predatory pricing
ONDC and DPDP-compliant audit trails from day one

The Commerce Dark Pattern Taxonomy

When agents operate autonomously, dark patterns scale at machine speed. The Consumer Protection Act 2019 and E-Commerce Rules prohibit these practices — but most platforms cannot detect them in their own agents:

Pattern Type	Agent Behaviour	Consumer Impact
Surge Pricing Exploitation	Pricing agent inflates prices during demand spikes, weather events, or supply disruptions without guardrails	Consumers pay 2-5x more with no transparency on pricing logic
Recommendation Manipulation	Agent prioritises high-margin products over relevance, buries competitive options in search rankings	Consumers see curated results that serve the platform, not their needs
Deceptive UI Agents	Customer service agent creates false urgency, fabricates stock scarcity, or misrepresents delivery timelines	Pressured purchases, inflated expectations, trust erosion
Unfair Ranking	Search and discovery agents favour platform-owned brands or paying sellers without disclosure	Smaller sellers disadvantaged; consumers denied fair comparison
Privacy Violations	Personalisation agents use browsing, location, and purchase history without adequate consent	DPDP Act violations; consumers profiled without knowledge or control

The Dark Pattern Problem

When a human salesperson uses a dark pattern, it affects one customer. When an agent encodes manipulative behaviour, it affects every interaction at production scale. Dark pattern prevention is an agent operations requirement.

Not generic recommendation engines adapted for India — autonomous agents solving the specific problems Indian retailers face every day. Each use case demands agent governance built for the Indian commerce context.

1. Vernacular Commerce Agents

India's next 300 million online shoppers are vernacular-first. They search in Hindi, negotiate in Tamil, compare prices in Telugu, and chat in code-mixed Hinglish. These shoppers are coming from Tier 2, 3, and 4 cities — and the agents serving them must do far more than translate. They must understand regional preferences, local festivals and shopping seasons, regional sizing conventions, and cultural context around gifting and occasions.

When a vernacular commerce agent recommends a product or declines a return, its reasoning must be explainable — not just in the developer's logs, but in the customer's language. Fairness monitoring ensures recommendation quality doesn't degrade for users interacting in less-resourced languages. Vishwas monitors agent recommendations for fairness across language, geography, and demographic categories. Orchestrate manages agent lifecycle, policy enforcement, and reasoning capture across every vernacular interaction.

2. Pricing and Inventory Agents

Quick commerce — Blinkit, Zepto, Instamart — grew over 60% year-on-year and demands real-time pricing decisions. A pricing agent operating across thousands of dark stores must decide in milliseconds: adjust prices for demand spikes, manage markdowns on perishables, coordinate stock transfers between locations. These agents operate with near-zero human oversight at 10-minute delivery speed.

Pricing agents without guardrails are a liability. Surge pricing that exploits demand spikes during extreme weather or supply disruptions is a regulatory and reputational risk. Guardian monitors pricing and inventory agents for drift, anomalies, and policy violations in real time. Orchestrate enforces bounded autonomy — guardrails that prevent predatory pricing while allowing agents to operate at the speed quick commerce demands.

3. ONDC Commerce Agents

ONDC is India's public digital infrastructure play for commerce — an open protocol where agents from different sellers, logistics providers, and buyer platforms interact without a central platform controlling the experience. A buyer agent on one app discovers products listed by a seller agent on another, with a logistics agent from a third provider coordinating delivery. This is multi-agent commerce at national scale.

Agent governance across an open network is fundamentally different from governance within a single platform. Every agent needs a verifiable identity. Policy enforcement must happen at the protocol level. Orchestrate provides agent identity, registration, and policy enforcement across ONDC's open network. Dastavez deploys document AI agents for invoicing, returns processing, and compliance documentation — with every agent action auditable against ONDC protocols and DPDP Act requirements.

The Common Thread

Every use case requires the same thing: agents that can be registered, monitored, explained, and bounded. The use case is specific. The governance architecture is universal.

Everyone focuses on cost-per-token. The right metric for commerce agents is cost-per-decision. And the 10x cost differences come from agent routing architecture, not provider negotiations. The Trust Cascade routes each agent decision to the cheapest sufficient intelligence layer.

"~65% of commerce agent decisions can be handled by rules. ~20% by traditional ML. Only ~15% genuinely benefit from agent reasoning. But most deployments route 100% through agents. That's not strategy — that's waste."

Agent Decision Routing: The Trust Cascade for Commerce

Layer	Volume (10L)	Cost/Decision	Monthly Cost
L1: Rules Engine (~65%)	6,50,000	Rs 0.001	Rs 650
L2: Statistical ML (~20%)	2,00,000	Rs 0.01	Rs 2,000
L3: Single Agent (~10%)	1,00,000	Rs 0.80	Rs 80,000
L4: Multi-Agent Tribunal (~5%)	50,000	Rs 3.50	Rs 1,75,000
Cascaded Total	10,00,000	Rs 0.26 avg	Rs 2,57,650
Pure Agentic (all LLM)	10,00,000	Rs 3-12	Rs 30L-1.2Cr

The Six Architectural Sins of Commerce Agent Deployment

1. Monolithic Prompts

Full product catalogue context on every recommendation call. You're paying for tokens the agent ignores on simple searches.

2. Retrieval Firehose

Stuffing all product matches into agent context. Your RAG retrieves 50 products when the answer requires three.

3. Retry Spiral

35% of agent requests involve retries. That's 35% cost overhead plus latency that kills the 10-minute delivery SLA.

4. Context Amnesia

No semantic caching. Same product query from 1,000 shoppers = 1,000 identical inference costs. No shared reasoning.

5. One-Agent-Fits-All

Frontier models for everything, including "is this product in stock" queries a rules engine handles perfectly.

6. Verbose Agent Output

Agent asked for a price recommendation, responded with a marketing essay. Output tokens cost 3-4x input tokens.

The Multiplier Effect

These sins multiply: 2x (monolithic) x 1.5x (firehose) x 1.35x (retry) x 1.4x (no cache) x 1.5x (verbose) = 8.5x optimal cost. Agent operations architecture eliminates this waste.

India's regulatory landscape for e-commerce is tightening. The Consumer Protection Act 2019, E-Commerce Rules 2020, DPDP Act, and ONDC protocols collectively demand agent governance that most retailers haven't built. When your agents are autonomous, every consumer protection requirement becomes an agent operations requirement.

Regulatory Requirements Mapped to Agent Governance

Regulation	Requirement	Agent Governance Need	Rotavision
CPA 2019	Prohibition of unfair trade practices	Dark pattern detection and prevention in agent behaviour at runtime	Guardian
CPA 2019	Product liability and misleading claims	Hallucination detection for customer service agents making product claims	Guardian
E-Commerce Rules	Price manipulation prohibition	Pricing agent guardrails with audit trail for every price change	Orchestrate
E-Commerce Rules	Search ranking transparency	Reasoning capture for recommendation and ranking agent decisions	AgentOps
DPDP Act	Consent and data minimisation	Personalisation agents operate within consent boundaries, data access policies	Sankalp
DPDP Act	Right to erasure and data portability	Agent reasoning traces must be deletable; personalisation models retrainable	AgentOps
ONDC Protocol	Open network interoperability	Agent identity, cross-platform policy enforcement, protocol-level governance	Orchestrate

ONDC Compliance Architecture

Agent Governance for Open Network Commerce

Identity

Every network participant has verifiable identity

Agent registry with verifiable identity across buyer, seller, and logistics platforms

Most agents operate without identity

Policy

Protocol-level rules for fair commerce

Policy engine enforced at Beckn protocol layer, not just within each platform

Policies enforced manually, post-hoc

Audit

Transaction traceability across network

Reasoning capture for cross-platform agent interactions and dispute resolution

Basic transaction logs only

The Compliance Reality

Consumer protection isn't a legal checkbox — it's an agent governance requirement. If your pricing agent can't explain why it charged what it charged, you cannot comply with the Consumer Protection Act 2019.

Deploying a commerce agent is not the same as operating one. The Agent Operations Stack is the infrastructure layer between your agents and production — ensuring every pricing agent, recommendation agent, and customer service agent is registered, governed, monitored, and bounded before it touches a single customer.

"The industry doesn't have a commerce agent deployment problem. It has an agent operations problem. The agents work. The infrastructure to govern them doesn't exist."

Five Layers of Commerce Agent Operations

1

Agent Registry

Every commerce agent registered with a unique identity, version, owner, and consumer impact scope. Pricing agents, recommendation agents, customer service agents — each classified by autonomy level and risk tier. No agent operates in production without registration. The single source of truth for what agents serve your customers.

2

Policy Engine

Dark pattern prevention rules enforced at runtime. Pricing guardrails that cap surge pricing. Recommendation fairness policies across demographics. Customer service hallucination boundaries. Policy as code — version-controlled, auditable, and enforceable at gateway, sidecar, and inline layers.

3

Reasoning Capture

The flight recorder for commerce decisions. Every pricing change, product recommendation, customer interaction, and inventory decision captured with full provenance. When a consumer asks why they were charged a specific price, you have the complete trace — not a log file, but a reconstructable decision path.

4

Bounded Autonomy

Pricing agents operate within guardrails — surge pricing limits, minimum margin floors, markdown thresholds. Recommendation agents stay within fairness boundaries. Customer service agents escalate when uncertain rather than hallucinate. The boundaries are configurable per agent, per channel, per risk tier.

5

Human-in-the-Loop

Not a checkbox — a workflow. When pricing agents detect anomalous demand patterns, humans review before prices spike. When customer service agents face complex complaints, the full reasoning chain and confidence assessment are passed to human agents. Escalation decisions are logged and learned from.

The Rotavision Difference

Operations, not just deployment. Every layer is built for Indian commerce — where an ungoverned agent isn't just an engineering risk, it's a consumer protection violation.

Before any commerce agent launches in production, it must clear five gates. These aren't bureaucratic hurdles — they're the foundations of agent operations that will satisfy consumer protection regulators and keep your systems reliable at the scale Indian e-commerce demands.

1

Gate 1: Agent Registration

Agent registered in enterprise registry with unique identity, version, owner, and consumer impact scope. Data access boundaries defined — what customer data the agent can use for personalisation. Pricing boundaries documented — maximum markups, minimum margins, surge caps. No unregistered agents in production.

2

Gate 2: Reasoning Capture

Flight recorder active for every agent decision. Every recommendation, pricing change, and customer interaction stored with full reasoning chain. Why was this product shown first? Why was this price set? Why was this return declined? Complete trace reconstructable for any historical decision.

3

Gate 3: Reliability Monitoring

Drift detection enabled for recommendation quality over time. Hallucination detection active for customer service agents — catching fabricated product features, false delivery promises, and invented policies. Anomaly detection for pricing agents. Alerts configured with on-call routing for production incidents.

4

Gate 4: Fairness Monitoring

Recommendation bias monitoring across languages — ensuring quality doesn't degrade for Tamil, Telugu, or Bengali users compared to English and Hindi. Geographic fairness across Tier 1 versus Tier 2/3/4 cities. Demographic monitoring to prevent discriminatory pricing or recommendation patterns. Continuous monitoring, not one-time testing.

5

Gate 5: Bounded Autonomy

Pricing guardrails configured and tested — surge caps, margin floors, markdown limits. Dark pattern prevention rules enforced at runtime. Human-in-the-loop workflows active for anomalous pricing decisions and escalated customer complaints. Cost controls and rate limits operational. Graceful degradation to lower-cost layers defined.

"A commerce agent should not launch until all five gates are cleared. In Indian retail, this isn't optional — it's the minimum bar for agent operations and Consumer Protection Act compliance."

Rotavision provides the complete agent governance infrastructure for Indian retail and e-commerce. Six products built from first principles for commerce agent operations, Indian languages, consumer protection, and ONDC compliance.

Orchestrate

Multi-Agent Commerce Operations

Enterprise-grade agent orchestration with Trust Cascade routing, policy enforcement, and bounded autonomy. Agent registry, reasoning capture, and human-in-the-loop workflows. The operational backbone for governed agent deployment across marketplaces, quick commerce, and D2C channels.

Vishwas

Agent Fairness for Pricing and Recommendations

Fairness monitoring built for Indian commerce — detecting recommendation bias across languages, geographies, and demographics. Pricing fairness across customer segments. Consumer-facing explainability in 22 languages. Consumer Protection Act 2019 alignment.

Guardian

Agent Reliability Monitoring

Continuous production monitoring for commerce agent behaviour. Catches drift in recommendation quality, hallucination in customer service agents, and anomalies in pricing agents. Dark pattern detection at runtime. 96% detection accuracy at less than 50ms overhead.

Dastavez

Document AI and Browser Agents

Document AI agents for invoicing, returns processing, and compliance documentation across ONDC's open network. Browser agents for competitive intelligence and catalogue management. Every agent action auditable against ONDC protocols and DPDP Act requirements.

AgentOps

Enterprise Agent Registry and Policy Engine

From RotaScale. Centralised agent registry with identity, autonomy levels, and consumer impact classification. Declarative policy engine enforced at runtime. Flight recorder for every agent decision. The control plane for enterprise commerce agent operations.

Sankalp

Sovereign AI Gateway

Route agent traffic through India-hosted infrastructure with trust monitoring built in. Customer data never leaves India. AWS Mumbai, Azure India, private cloud, or hybrid deployment. DPDP Act compliance by design for all commerce agent traffic.

Built for Indian Commerce. ONDC-Ready. DPDP-Compliant.

Your infrastructure. On-premise, private cloud, or hybrid. No customer data leaves India. Every product built for agent governance in Indian retail. Consumer Protection Act 2019 and DPDP Act compliant from day one.

Production agent systems processing decisions across recommendations, pricing, inventory, and customer service. Each implementation demonstrates what becomes possible when commerce agents have proper operations infrastructure.

Vernacular Recommendation Agent with Fairness Monitoring

Recommendation agents serving 22 languages across Tier 1 through Tier 4 cities. Vishwas monitors every recommendation for fairness across language, geography, and demographics. Full reasoning capture for why products are shown. Explainability in the customer's language.

Result: Recommendation quality parity across all 22 languages

Dynamic Pricing Agent with Surge Protection

Pricing agent managing real-time price adjustments across marketplace, quick commerce, and D2C channels. Bounded autonomy with surge caps, margin floors, and markdown thresholds. Every pricing decision traced with full reasoning. Consumer Protection Act compliant.

Result: 35% margin improvement with zero predatory pricing incidents

Quick Commerce Inventory Agent

Autonomous inventory agent coordinating stock across hundreds of dark stores. Demand forecasting, perishable management, and stock transfer decisions at 10-minute delivery speed. Guardian monitors for drift and anomalies. Escalation to human when demand patterns are anomalous.

Result: 28% reduction in stockouts, 40% less perishable waste

ONDC Cross-Platform Agent Governance

Agent governance across ONDC's open network. Verifiable agent identity for buyer, seller, and logistics agents. Cross-platform policy enforcement at the Beckn protocol layer. Reasoning capture for multi-agent transactions spanning multiple network participants.

Result: Full audit trail across open network agent interactions

Customer Service Agent with Hallucination Detection

Customer service agents handling queries, complaints, and returns in 22 languages. Guardian detects hallucinated product features, fabricated policies, and false delivery promises in real time. Bounded autonomy with escalation to human agents for complex complaints.

Result: 92% resolution rate with zero hallucination-related escalations

Dark Pattern Prevention Dashboard

Real-time monitoring for manipulative agent behaviours across all commerce agents. Surge pricing exploitation, deceptive recommendation patterns, unfair ranking, and false urgency creation detected and prevented. Consumer Protection Act 2019 compliance scoring.

Result: Continuous dark pattern monitoring across all agent touchpoints

"The platform doesn't replace your commerce strategy — it makes your agents production-ready for Indian consumer protection. Same capabilities, but with the governance infrastructure regulators expect."

A combined assessment, platform, and integration package for retailers deploying AI agents across pricing, recommendations, and customer service — with Consumer Protection Act 2019 and DPDP Act compliance built in.

What's Included

1

Commerce Agent Trust Assessment

Audit agent readiness across pricing, recommendation, and customer service agents against Consumer Protection Act 2019 and DPDP Act requirements. Gap analysis with compliance roadmap and prioritised remediation plan.

2

Agent Registry for Commerce Operations

Orchestrate + AgentOps configured for retail — pricing agents, recommendation agents, and customer service agents registered with autonomy boundaries and consumer impact scope. Policy enforcement at runtime.

3

ONDC and Beckn Protocol Integration

Agent governance for the open network — verifiable agent identity, cross-platform policy enforcement, and reasoning capture across ONDC's decentralised commerce protocol. Beckn-compliant agent registration.

4

Dark Pattern Prevention Engine

Continuous monitoring for manipulative agent behaviours — surge pricing exploitation, deceptive recommendation patterns, and unfair ranking algorithms. Consumer Protection Act 2019 compliant. Real-time prevention, not post-hoc detection.

5

Consumer-Facing Explainability

When agents recommend, price, or decline — the customer gets a vernacular explanation. Reasoning surfaced through existing app interfaces in 22 languages. DPDP-compliant consent capture at the agent level.

Platform Stack

Agent orchestration

Orchestrate

Fairness and explainability

Vishwas

Reliability monitoring

Guardian

Document AI

Dastavez

Agent registry and policy

AgentOps (RotaScale)

Sovereign gateway

Sankalp

Engagement Options

Assessment

Rs 12L

2 weeks. Commerce agent trust audit. Consumer protection gap analysis. Dark pattern assessment. Board-ready roadmap.

Accelerator

Rs 25L

4 weeks. Full compliance alignment. Agent registry setup. Dark pattern prevention. Executive presentation.

Production

Rs 40L+

8-16 weeks. Full platform deployment. ONDC integration. Team training. Go-live support.

Rotavision India

India's next 300 million shoppers speak vernacular. The agents serving them need governance, not just good recommendations.

500 million online shoppers by 2030. Pricing agents deciding what they pay. Recommendation agents deciding what they see. Customer service agents deciding what they're told. ONDC creating multi-agent commerce at national scale. The Consumer Protection Act demands accountability that most platforms haven't built. The agents are already deployed. The operations layer is what's missing.

We'd like to show you where you stand. A 30-minute assessment — not a sales pitch — to benchmark your agent governance against consumer protection requirements and identify your highest-value opportunities.

Request Assessment

Agent Governance for Indian Commerce

Four Gaps That Define the Problem

The Recommendation Agent Gap

The Pricing Agent Gap

The Quick Commerce Gap

The ONDC Governance Gap

Commerce Agents Without Operations

Commerce Agents with Rotavision

The Commerce Dark Pattern Taxonomy

1. Vernacular Commerce Agents

2. Pricing and Inventory Agents

3. ONDC Commerce Agents

Agent Decision Routing: The Trust Cascade for Commerce

The Six Architectural Sins of Commerce Agent Deployment

1. Monolithic Prompts

2. Retrieval Firehose

3. Retry Spiral

4. Context Amnesia

5. One-Agent-Fits-All

6. Verbose Agent Output

Regulatory Requirements Mapped to Agent Governance

ONDC Compliance Architecture

Agent Governance for Open Network Commerce

Five Layers of Commerce Agent Operations

Agent Registry

Policy Engine

Reasoning Capture

Bounded Autonomy

Human-in-the-Loop

Gate 1: Agent Registration

Gate 2: Reasoning Capture

Gate 3: Reliability Monitoring

Gate 4: Fairness Monitoring

Gate 5: Bounded Autonomy

Orchestrate

Vishwas

Guardian

Dastavez

AgentOps

Sankalp

Vernacular Recommendation Agent with Fairness Monitoring

Dynamic Pricing Agent with Surge Protection

Quick Commerce Inventory Agent

ONDC Cross-Platform Agent Governance

Customer Service Agent with Hallucination Detection

Dark Pattern Prevention Dashboard

What's Included

Commerce Agent Trust Assessment

Agent Registry for Commerce Operations

ONDC and Beckn Protocol Integration

Dark Pattern Prevention Engine

Consumer-Facing Explainability

Platform Stack

Agent orchestration

Fairness and explainability

Reliability monitoring

Document AI

Agent registry and policy

Sovereign gateway

Engagement Options

India's next 300 million shoppers speak vernacular. The agents serving them need governance, not just good recommendations.

Rotavision Consulting Private Limited