- Set Clear Rules for AI Agents: As retailers let AI agents handle more tasks, they need to define what agents can do, when people must step in, and how to fix mistakes.
- Give Your Internal Agents Accurate Information: Retailers’ internal AI agents need up-to-date product details, stock levels, and store policies to fill orders correctly and avoid promises the retailer can’t keep.
- Keep Customers in Charge: Before a consumer agent makes a purchase or changes an order, check that the customer gave permission to that agent and that their action stays within their limits.
- Track Whether Agentic Commerce Is Working for Customers: Measure whether agents follow customer instructions, how often purchases lead to disputes or need to be undone, and whether customers can still deal directly with the retailer.
Agentic Commerce Moves Retail Identity From Access to Agency
In 2027, leading retailers will be deciding where to let AI agents act, how much authority to give them, and how to stay accountable for the results.
Agentic commerce is moving beyond chatbot demos into discovery, comparison, basket building, service, payments, store operations, and supply chain workflows.
But the market isn’t moving toward AI agents making every purchase on their own. Instead, people are giving agents more tasks while setting clear limits on what they can do, when they need approval, and how mistakes can be corrected.
The strategic question is no longer whether AI can participate in commerce. It is which parts of the retail journey should be delegated to a particular agent, on whose terms, and how retailers can confirm that the agent acts within those terms.
What Is Agentic Commerce?
Agentic commerce is when a customer gives an AI agent permission to handle shopping tasks on their behalf. The agent can find products, compare options, build a basket, make purchases, and arrange returns within the customer’s instructions, budget, and approval requirements.
This is in contrast to AI-assisted shopping where AI offers suggestions and answers questions, while the customer chooses what to buy and completes the purchase.
How AI Agents Are Becoming a Retail Distribution Layer
AI agents are becoming a factor in how retailers plan for growth. McKinsey estimates that agents could help shape or carry out $3 trillion to $5 trillion in global consumer purchases by 2030, even under moderate scenarios.¹
For now, shoppers are using AI more to find and compare products than to complete purchases. According to Checkout.com, US and UK merchants estimated that only 3% of their transactions involved agents, while 89% of merchants overall were preparing for agent-led shopping.²
Retailers therefore need to prepare for more than fully automated checkout. An AI assistant may influence which products a shopper considers before they ever visit a retailer’s website or app. The shopper might then complete the purchase themselves or give an agent permission to do it.
When an agent reaches the retailer, that distinction matters. Is it gathering product information, preparing a basket for review, or placing an approved order? Each task needs different permissions.
Retailers need to make their products easy for agents to find and understand, while checking permission before agents access customer information or take action. Identity controls help them recognize the agent, confirm whom it represents, and check what it is allowed to do.
How Much Will Shoppers Let AI Agents Do in 2027?
Consumers are adopting AI shopping assistance faster than they are trusting AI to make payments. Visa found that 72% of surveyed consumers had used an AI assistant, but only 23% of US consumers trusted generative AI to handle their payments.³
Shoppers also want ways to stay in control. Checkout.com found that consumers’ top non-negotiables for agent shopping include spending caps (30%), the ability to withdraw permission immediately (29%), and easy cancellation (28%).²
For retailers, this means 2027 should be about setting up support for five different levels of customer control across the shopping journey:
- Finding products: An agent searches for items and explains the options. The customer decides what to do next.
- Choosing what to buy: An agent compares prices, features, and availability, then recommends a product. The customer makes the final choice.
- Getting an order ready: An agent adds items to a basket and prepares checkout. The customer reviews and approves the purchase.
- Making an approved purchase: A customer lets an agent buy within specific limits. For example, re-ordering the same household supplies once per month if under $50.
- Handling tasks after purchase: A customer lets an agent track a delivery, arrange a return, or request a refund. Changes outside the customer’s instructions require approval.
Customers may allow agents to do some of these tasks while keeping others for themselves. Someone might let an agent reorder coffee, but require approval for substitutions. Another shopper might approve a purchase, but wants to be asked again if the delivery address changes.
Retailers need to make those choices easy to set, change, and cancel. Before accepting an order or a change from an agent, they need to check who gave permission, what the agent is allowed to do, and whether that permission is still valid. Customers also need a clear way to reach the retailer and resolve mistakes.
Why Product Data Decides Which Retailer an AI Agents Chooses
AI shopping agents rely on product information to compare options: features, price, stock levels, delivery dates, reviews, loyalty benefits, and return policies. If that information is missing or inconsistent, an agent may overlook a retailer’s products or give a shopper incorrect expectations.
Retailers need to keep that information accurate and up to date wherever agents find it. A shared source of product, inventory, and policy information can help prevent different shopping channels from showing conflicting details.
This makes accurate information essential to winning and fulfilling an order. An agent needs to know whether a product meets the shopper’s needs, whether it is available, and whether the retailer can deliver it when promised.
Customer-specific information requires another check. Any agent may be able to look up a public price or return policy. But before it can use a customer’s loyalty benefits, view an order, or request a return, the retailer needs to confirm which customer it represents and whether that customer has given permission.
Reliable product information helps agents make useful recommendations. Clear customer permissions help retailers turn those recommendations into purchases and service customers can trust.
Different Agents Need Different Rules
AI agents play different roles across retail. Some help employees work, some run business processes, some act for shoppers or partners, and others operate inside vendor software. Retailers need to understand each agent’s role and set appropriate limits.
- Consumer and partner agents act on someone else’s behalf. A shopper’s agent might place an order, while a supplier’s agent might update a listing. Before accepting the action, the retailer needs to check whom the agent represents and whether that person or business has authorized it. Customer benefits, such as loyalty rewards, should still apply where appropriate.
- Employee agents help staff get work done. An employee might use an agent to research pricing, review product listings, or answer a policy question. Retailers need to know when these tools are being used, protect sensitive information, and decide which actions require a person’s review.
- Workload agents carry out operational tasks. These agents may coordinate inventory, process returns, or support order fulfillment. Retailers need to assign someone responsibility for each agent, limit what it can change, and define when it must ask a person to step in.
- SaaS-embedded agents work inside software the retailer uses. An agent in a customer-service or warehouse platform may access retail data or trigger changes. Even when the vendor provides the agent, the retailer needs to control what it can access and keep a record of what it does.
Across all four types, the practical questions are the same: Who is the agent working for? What is it allowed to do? When does a person need to approve? And who can resolve a mistake?
Know Who an Agent Works for and What It Can Do
Before accepting an action from an AI agent, a retailer needs to answer two questions: Who is the agent working for, and does it have permission to do this?
Knowing which customer account is involved is only the start. A shopper might let an agent compare products and build a basket but still want to approve the purchase. They might set a spending limit, allow delivery only to a specific address, or withdraw permission before the agent places the order.
The retailer needs to check those instructions before accepting the purchase or change. Permission to help with one task should also stay limited to that task. An agent arranging a return, for example, should have access to the order details it needs, rather than the customer’s entire account.
This is what delegated authority means in practice: a customer lets an agent carry out specific tasks on their behalf. The retailer should be able to tell which actions came from the customer and which came from the agent, what permission applied, and whether it was still valid.
Similar checks apply to other retail agents. A supplier’s agent may be allowed to update its own listings. An employee’s assistant may be allowed to look up information but need approval to change it. An inventory agent may flag a shortage but need a manager’s approval to move stock. An agent built into vendor software should have access only to the retail data and systems needed for its job.
Identity controls help retailers put these rules into practice. They help identify the agent, connect it to the person or business it works for, check what it is allowed to do, and record its actions. That gives retailers a way to support useful automation while keeping customers and staff in control.
AI Shopping Depends on What Happens Behind the Scenes
Retailers need to prepare for AI agents on both sides of a purchase: external agents shopping for customers and internal agents helping run the business.
External agents can help shoppers find and buy products. Internal agents help retailers deliver on those purchases—from keeping product information accurate to getting orders out the door and handling returns.
For retailers, internal tasks may be an easier place to start. A merchandising team can use an agent to find missing product details. A fulfillment team can use one to flag orders at risk of delay. A customer-service team can use one to review order histories and prepare responses. Each task has a responsible team and a result the retailer can measure.
The benefits reach customers, too. More accurate stock information helps prevent orders for unavailable items. Earlier warnings about delays give staff time to respond. Faster access to order details helps service teams resolve problems.
But each internal agent needs clear limits. An inventory agent might check stock across stores but need a manager’s approval before moving it between locations. A service agent might explain a return policy but need approval before issuing a refund above a set amount.
Retailers should assign a person or team responsibility for each agent, give it access only to the information and systems it needs, and check that its actions stay within those limits. Keeping a record of its work helps staff understand what happened and correct mistakes.
This matters because a single purchase can involve the website, a warehouse, a store, a delivery provider, and customer service. Internal agents can help coordinate that work, but staff need a clear way to step in when an item is unavailable, a delivery is delayed, or a customer changes an order.
An external agent may place the order. The retailer’s people, systems, and internal agents still have to fulfill the promise.
Five Decisions Retailers Should Make as They Expand AI Use in 2027
Before giving AI agents more tasks, retailers need to decide where agents can help, what they can do, and who is responsible when something goes wrong.
- Where will we use AI agents? Decide whether to support customers’ shopping agents, use internal agents to help employees and run operations, offer your own shopping assistant, or combine these approaches.
- What information can agents rely on? Identify how agents will get accurate product details, stock levels, prices, delivery options, and policies—and what they should do when information is missing or out of date.
- What can agents do without asking? Set limits for actions such as placing orders, changing prices, moving inventory, or issuing refunds. Specify when a customer or employee must approve.
- How will we check whom an agent is working for? For a shopper’s agent, confirm the customer’s permission and apply the right loyalty benefits. For an internal agent, assign a responsible person or team and limit which systems and information it can use.
- Who will fix mistakes? Decide who can review what an agent did, cancel or correct the action, help the customer, and handle any resulting costs.
These decisions give each agent a clear job, clear limits, and someone responsible for its work. They also help retailers choose useful starting points and measure results before expanding.
Build Trust Into Every Agent Action for Retail Advantage in 2027
As AI agents take on more shopping and operational tasks, retailers need to make sure those agents follow instructions, use accurate information, and stay within clear limits.
For customers, that means knowing an agent will respect their budget, ask for approval when needed, and leave them a way to correct mistakes. For retail teams, it means agents have reliable product and inventory information, access only to the systems they need, and a person or team responsible for their work.
Ping Identity helps retailers identify agents, confirm whom they work for, and control what they can access and do. Those controls help retailers check permissions before an action takes place and keep a record of agent activity so teams can investigate problems.
The retail advantage will come from giving agents useful work and making sure they do it reliably. Whether an agent is placing an order for a shopper or helping a retailer fulfill it, the goal is the same: keep the promise to the customer.