Service & returns
- Resolves order status, returns, and product questions grounded in your policies — refunds execute instantly under the threshold and pend for approval above it.
- Action Fabric
AI agents for retail & e-commerce
Service, returns, pricing, and demand — with refunds that can't run away.
Retail agents resolve customer requests, manage returns, and support merchandising — but money-moving actions are governed objects: a refund over your threshold pends for approval, payment data stays locked, and every promotion decision carries its policy trail.
Aligned with PCI DSS · CCPA/CPRA · GDPR · SOC 2
Action Fabric
Dry-run›
Refund $240
risk: low · evidence
Proposed›
Close case #4471
risk: med · evidence
Pending approval›
Payout $5,000
risk: high · evidence
Executed
Email customer
risk: low · evidence
actions ▸ dry-run → propose → approve → execute → compensate
The regulatory pressure
The fastest way to lose a retail brand is an agent that refunds by vibes.
Customer-facing automation touches money, payment data, and pricing integrity at consumer scale. PCI DSS fences card data, CCPA/CPRA and GDPR fence shopper PII, and the FTC watches every claim your pricing engine makes. Cortex turns refunds, price changes, and promotions into governed actions with thresholds, approvals, and a sealed record.
PCI DSSCCPA / CPRAGDPRSOC 2FTC ActNIST AI RMF
Regulatory pressure
PCI DSSCCPA / CPRAGDPRSOC 2FTC ActNIST AI RMF
Typed objects
Restrictedlocked
frameworks ▸ encoded as policy · gated · sealed
Use cases
Every retail role, from service desk to demand planning.
The catalog ships 12 retail roles × 18 agent patterns — service, merchandising, inventory, loss prevention — each wired to your commerce stack and run through the same gates.
Pricing & promotions
- Drafts price changes and promotion plans checked against policy-as-code before anything ships — margin floors and claim rules enforced, not suggested.
- Policy-as-Code
Inventory & demand
- Forecasts demand and proposes replenishment against typed Product and Order objects, under hard 30-day budget caps for the compute.
- Cost governance
Loss prevention
- Watches order and return streams for anomaly patterns and raises cases — detection & response can pause an abused flow automatically.
- Agent D&R
Payment-data isolation
- Card PANs and payment tokens are locked deny-by-default at the property level — service agents help the customer without ever reading the card.
- Ontology permissions
Governed refunds
The agent can help a customer — without being able to empty the till.
Refunds and appeasements are first-class governed actions with risk tiers. Under your threshold they execute and seal instantly; above it they pend on the approval board with full context. Payment data never enters the loop — the property is locked before the agent ever runs.
- Refund threshold routes big appeasements to human approval
- payment_token / card_pan locked deny-by-default (403 on read)
- Promotion spend bounded by hard budget caps (402 at 100%)
- Every refund, price change, and promo sealed in the ledger
Action Fabric
Dry-run›
Refund $240
risk: low · evidence
Proposed›
Close case #4471
risk: med · evidence
Pending approval›
Payout $5,000
risk: high · evidence
Executed
Email customer
risk: low · evidence
actions ▸ dry-run → propose → approve → execute → compensate
Prove it — don't just claim it
The verdicts your commerce lead will actually see.
Literal runtime responses — money-moving actions fail closed.
runtime · verdicts
Refund $49200 sealed
Refund $2,500HOLD · approval
Card PAN read403 denied
Promo overspend402 budget
fail-closed ▸ challenge it and it denies, on the record
- 01
Model the domain
Map the work to typed ontology objects — Order, Product, Customer, Return, Promotion — with restricted fields like payment_token / card_pan locked at the property level.
- 02
Encode the rules
Translate the obligations above into Policy-as-Code and Action Fabric approvals — the gates fail closed, returning a precise code, not a guess.
- 03
Prove the outcome
Every run lands in a hash-chained Trust Ledger with signed receipts and 10-hop provenance — a record you can hand the regulator.
100%Refunds recorded — executed, held, or blocked
8 gatesOn every customer-facing run
10-hopProvenance on every decision
5 modesAutonomy dial per agent and action
Works with the rest of the runtime
The controls behind retail & e-commerce.
The same shared runtime powers every vertical — here are the capabilities and the Solution Pack this industry composes from.
Action Fabric
Refunds as governed actions: dry-run, approve, execute, compensate.
Policy-as-Code
Margin floors and promo rules as testable, simulated policy.
Cost governance
Hard caps on forecasting and personalization compute.
Ontology
Model Order, Return, Promotion; lock payment properties.
Agent Library
12 retail roles × 18 patterns on Shopify, Salesforce Commerce, NetSuite.
Trust Center
Map payment isolation and consumer-data handling to PCI and CPRA.
Security & compliance
Consumer scale needs controls that don't blink.
Payment isolation, refund thresholds, DLP on every customer conversation, and detection & response are shared across service, merchandising, and analytics — mapped once to PCI DSS and CPRA obligations. Aligned with, never claiming a certification you don't hold.
PCI DSS · scopeCCPA / CPRAGDPR · consumerSOC 2 · auditFTC · claimsNIST AI RMF · Manage
FAQ
AI agents for retail & e-commerce — questions, answered.
Can an AI agent issue refunds on its own?
Only within the threshold you set. Refunds are governed actions with risk tiers — under the threshold they execute and seal instantly; above it they pend on an approval board with full order context. Either way the ledger records the outcome.
How does Cortex keep agents away from payment data?
Card PANs and payment tokens are locked deny-by-default at the ontology property level, so a service agent reads the Order and the Return but a payment-data read returns a 403 — before any model is invoked.
Can agents change prices or launch promotions?
They can draft them — every price change and promotion is checked against policy-as-code (margin floors, claim rules, spend caps) and routes to approval when it crosses a threshold. Simulation shows what a policy would do before you ship it.
What retail agents are in the library?
Twelve retail roles — merchandiser, category manager, pricing analyst, inventory planner, e-commerce manager, store operations, loss prevention and more — each in 18 agent patterns, wired for systems like Shopify, Salesforce Commerce Cloud, SAP Retail, NetSuite, and Square.
More industries
One governed runtime, every regulated vertical.
The controls are the same — the obligations they satisfy differ. Explore another vertical, or see the full set.
Automate the storefront — keep your hands on the till.
See governed refunds, locked payment data, and policy-checked pricing on your own commerce stack.