.gif){kind=spacer}
TL;DR
When a delivery runs late, the customer is at home. When a BOPIS order fails, the customer is standing at your counter, or waiting in your parking lot, for an item that isn't there.
Most of those failures are decided before the customer leaves home: the store sold stock it didn't have, nobody asked about the substitute, or "ready for pickup" went out before the order was packed. AI agents catch them there. They triage problem orders by pickup deadline, get the customer's approval on substitutes, and confirm readiness only when the order is staged. After packing, they alert associates when a curbside customer arrives and clear uncollected orders back to the shelf.
Start with exception triage and substitution approval. That is where order fill rate and pickup satisfaction move first.
Why does BOPIS put so much strain on store teams?
Buy online, pick up in store (BOPIS), also called click and collect, is now a mainstream way to shop, and every order lands on the store floor. 68% of US shoppers have used it at least once, up from 57% in 2022.
The upside is real foot traffic. 49% of customers made additional purchases when they picked up in store. Half of shoppers have chosen where to buy online based on whether in-store pickup was available.
The catch is speed. A quarter of BOPIS shoppers want to collect within two to four hours. That means associates pick online orders from the same shelves walk-in customers shop, in the same shift, while also running the register and answering the store phone.
Every order passes through five moments after it's placed. Each one can break:
- Picking: an associate finds each item on the shelf.
- Substitutions: a missing item needs a replacement the customer will accept.
- Ready for pickup: the customer is told, and the pickup window starts.
- Handover: at the counter or at the curb.
- Uncollected orders: reminders, cancellation and restocking.
This post goes deep on one workflow, the BOPIS order. For the wider picture of how AI agents support store teams, see our guide to AI agents for in-store associates.
Why do BOPIS orders go out of stock at picking?
Most BOPIS exceptions start before anyone touches the shelf: the website sells stock the store doesn't have. Researchers at Auburn University note that BOPIS out-of-stocks happen because store inventory records are not always accurate. The system shows three units; the shelf shows none.
So the associate arrives at the bay and finds one of three things:
- An empty shelf: the record says three units, the shelf says zero.
- A partial order: four of five items found, one missing.
- The wrong item: a different size, color or pack count sits where the SKU should be.
Today, each one becomes a judgment call made on the spot. The associate walks to a manager or a terminal, decides whether to substitute or cancel, and the order sits while the pickup clock runs.
How should AI handle BOPIS substitutions?
A good substitution depends on what was ordered, and a wrong one costs more than a cancellation. An Auburn University study published in the Journal of Business Logistics found BOPIS shoppers accept substitutes more readily for staples (ketchup, paper towels) than for indulgence items (coffee, perfume). The researchers concluded that highly indulgent categories may be poorly suited to substitution offers at all.
The same researchers point to AI as a way to find optimal substitutes. The industry view is to give customers options well before they travel to the store. Asking shoppers to pre-select substitutes for every item at checkout is tedious, so the better pattern is to ask only about items at real risk, or to send an approval request after checkout.
When should a click-and-collect customer hear their order is ready?
As soon as it is truly ready, and not a minute before. In a 2018 mystery-shopper study by IHL Group, shoppers notified within four hours were 19% more likely to use the service again and 15% more likely to recommend it. The same research found 26% of shoppers arrived to find their order not ready, even after being told it was.
A notification is a promise. When the message goes out before the order is staged, the store has turned a fulfillment problem into a customer-service problem.
Many "Is my order ready?" questions reach the store as phone calls. We cover how agents handle them in our guide to WISMO calls in stores.
How can AI speed up in-store and curbside pickup?
Handover is the last moment the customer judges the whole experience, and it often depends on an associate noticing them. One industry guide sets the target at under five minutes of pickup wait. Public data on actual wait times is thin, but the failure pattern is familiar: a curbside customer parks and waits because nobody saw the arrival, or a counter customer waits while an associate hunts the staging area.
What happens to BOPIS orders nobody collects?
An uncollected order is stock that walk-in shoppers can't buy, sitting in the staging area. Reliable public figures on uncollected-order rates are scarce, but the cost is easy to see on any busy pickup shelf: paid-for or held items off the sales floor, and staging space filling up.
One industry guide recommends a reminder if an order isn't picked up within 48 hours. The harder part is what follows: someone has to decide when to cancel, process the refund and walk every item back to its shelf.
What does an AI agent do at each stage of BOPIS order management?
The agent takes the lookups, decisions and messages off the associate, so the associate only picks, packs and hands over.
Which BOPIS metrics should store leaders track, and where should they start?
Track six numbers, one per break point plus resolution speed:
- Order fill rate: share of orders fulfilled as placed or with an approved substitute.
- Substitution acceptance rate: share of offered substitutes the customer approves.
- Time to resolve an exception: from the associate finding the problem to the order being updated.
- Orders ready on time: share staged before the promised pickup time.
- Pickup wait time: from the customer's arrival to the order in hand.
- Uncollected-order rate: share of orders canceled after the pickup window.
Where to start
Start with exception triage and substitution approval. They happen on every busy shift and their impact shows in weeks. Customer approval reduces substitution errors, but delays can still cost sales. Set a response deadline and a clear fallback before changing orders.
Add proactive notifications and curbside alerts next, once order data from the order management system is flowing reliably. Leave uncollected-order handling for last, because it depends on refund and cancellation policies being written down clearly enough for an agent to apply.
For how BOPIS ranks against other store-floor workflows, see the five in-store use cases to automate first.
How NuPlay store experts handle BOPIS
NuPlay's retail agents resolve 95% of store associate queries autonomously. For BOPIS, the work splits across four capabilities:
- Store audits and inventory checks: cycle counts reviewed against the retailer's standard, with exceptions routed and fixes tracked to close. This is what keeps phantom stock from selling twice.
- Order, return and loyalty resolution: substitutions, cancellations and refunds handled end to end, each decision citing the policy clause it applied.
- Proactive outreach: ready-for-pickup messages, delay alerts and reminders sent before the customer has to ask.
- Intelligent escalation: a missing bag or wrong order reaches a person with the history, intent and full customer context attached.
The agents work across voice, chat, SMS, email, app and messaging. They connect to POS, ERP and CRM systems and to commerce platforms including Shopify, Salesforce Commerce Cloud, Magento and Kibo. NuLoop, NuPlay's self-improvement layer, learns from every exception the agent resolves, so substitution suggestions get sharper with each order.
