# Conversational AI for retail: Order, return, and store calls

By Samuel Oram-Jones

Conversational AI for retail is software that answers a shopper's phone call in everyday language, remembers what was said earlier, and works through an order status check, a return, or a question about a store. Shoppers phone when something did not arrive or did not fit, and the question usually comes down to one fact.

Bland's agents are inbound and outbound AI phone agents, built and edited without code in Norm or Bland's visual builder.

## What does conversational AI do on a retailer's phone lines?

On a retailer's phone lines, conversational AI turns a shopper's sentence into a lookup and a plain answer. A shopper who says *my jacket never showed up* gets an order status, with no menu and no wait for a person to open a screen.

Inbound agents take order status calls, return requests, and questions about store hours and stock. Outbound agents place the calls a retailer starts, such as telling a shopper that an order is ready for pickup.

Silence on a phone line reads as a fault, so speed matters. Bland's agents run at sub-200ms latency.

For the general definition, see [what conversational AI is](/blog/conversational-ai). Bland's [conversational AI platform](/conversational-ai) page covers the product, and the [best conversational AI platforms](/blog/best-conversational-ai) list ranks the options.

## Which retail calls can a conversational AI agent take, and which stay with people?

An agent takes the retail calls that end in a lookup or a short fixed form, and people keep the calls where a shopper is angry, out of pocket, or asking for an exception.

| Call | Who takes it first | What decides it |
|---|---|---|
| Where is my order | AI agent | One order number and one lookup |
| Start a return | AI agent | An order number, an item, and a reason |
| Store hours or a store's stock | AI agent | A store's published details and a stock lookup |
| Pickup ready or delivery change, outbound | AI agent | One message and one confirmation |
| Refund dispute | Person | Authority to adjust money |
| Damaged item or safety complaint | Person | Judgment and care |
| Fraud claim on an order | Person | A wrong guess costs too much |

Order status is the call to automate first, because the shopper wants one answer and the answer sits in an order system. Returns come second, since a return is a form read aloud: the order, the item, and why it is going back.

What stays with people is the call where the shopper has already tried twice. An agent's job on that call is a clean handover, and Bland's Transfer Call node can send a caller to a DID, an extension, or a SIP URI on a customer's trunk.

## How does a retail agent handle a return and an order that changed?

A retail agent handles a return by collecting its details in a fixed order and handling a shopper's side questions without losing its place. Strict mode is a pathway's default and suits steps where the order matters, such as ordered data collection, so a return form belongs in a strict section.

Sample call (illustrative). This is an example of how a return call could go, written for this page. It is not a recording of a real call or customer.

- Agent: Thanks for calling. I'm an AI agent for the store, and I can connect you to a person whenever you ask. What's the order number?
- Caller: 4417. Wait, first, where's my other order? It was due Tuesday.
- Agent: Your other order shipped and is expected Friday. Do you want to go back to the return?
- Caller: Yes. It's the blue jacket, and it doesn't fit.
- Agent: Got it, a return for the blue jacket because of size. I can start that now, and anything about a refund amount goes to a person.

The shopper jumped to a second order and came back. After a global node answers an off-topic question, a Bland agent returns to the step it was on and carries on with the conversation. In Flex Mode, Flex Groups let a team flex some sections and keep others strict, which fits a return form inside a looser chat about products. Bland's [Flex Mode](https://docs.bland.ai/tutorials/flex-mode) page holds the detail.

Orders change between calls, so memory matters. When memory is on, Bland keeps rolling summaries, structured facts, and open action items for each contact. Open items are follow-ups such as one on an order delay, so a shopper who phones back about a late order is not asked to start over. Bland's [memory](https://docs.bland.ai/tutorials/memories) page covers the setting.

Memory also carries across channels. A contact keeps one shared history across its voice calls, and on Enterprise its SMS and iMessage threads join that history.

## How does conversational AI connect to a retailer's order and store systems?

Conversational AI connects to a retailer's order and store systems through Bland's REST API and webhooks. Salesforce, HubSpot, Cal.com, Calendly, Notion, and Slack connect from Bland's dashboard. Anything else connects through Bland's REST API and webhooks, which Bland's forward-deployed engineers can help with on Enterprise.

A webhook node executes webhooks at any point during a conversation, so an agent can look up an order before it tells a shopper where it is. A retailer with a store phone system keeps it, because Bland connects to a carrier or phone system over SIP.

Bland's [integrations](/integrations) page lists every dashboard connection. A retailer's team builds and edits the agent itself in plain English with [Norm](/norm), Bland's AI assistant.

## What does conversational AI for retail cost?

Bland costs $0.14 a minute on the self-serve Start plan. Enterprise pricing is custom, set with Bland's sales team. Phone is billed separately, on a customer's own carrier or Bland's at pass-through cost.

Enterprise adds native iMessage agents, warm transfers, a BAA, SSO, data residency, isolated infrastructure run by Bland, and forward-deployed engineers, alongside SMS and web chat.

Bland's [pricing](/pricing) page lists what each plan includes.

Bland's published result comes from Idaho Housing and Finance Association, which is not a retailer and replaced its phone tree on a busy inbound line. A retailer's order-status queue peaks differently, so size it before borrowing the number.

Idaho Housing and Finance Association saves $750,000 a year by replacing its IVR with Bland, which handles 4,000 inbound calls a day and cut average call time by 20%.

> "We save thousands of dollars a day."

Greg Blake, Chief Information Officer, Idaho Housing and Finance Association

## Frequently asked questions

### Can conversational AI tell a shopper where an order is?

Yes, when an agent can reach a retailer's order system. An order status call is one lookup and one answer, and Bland's webhooks run that lookup during the call.

### Can an agent start a return over the phone?

Yes, a return is a short run of details, and an agent can collect them in order. A person takes the refund disputes and any exception to a return policy.

### Is Bland only for outbound calls?

No. Bland's agents answer inbound calls and place outbound calls, so one retailer can take order calls and place pickup calls.

### What models does Bland use?

Bland self-hosts its models and uses no third-party model providers. Bland operates a closed-source model it developed itself.

### Does Bland have a no-code option?

Yes. Norm, Bland's AI assistant, builds voice agents using natural language descriptions, and Bland's dashboard includes a visual builder for non-technical users.

### How was this page put together?

Bland's features are described as Bland's documentation and signed fact file state them, captured on 20 November 2026. The sample call is written for this page and is not a recording, and the page uses no market statistics or retail data.
