WISMO calls start as a question about a record
A WISMO call asks where an order is. The caller wants a status and a date. The answer sits in your order system, and someone on your team opens it and reads the answer out.
Volume can follow your calendar. A promotion, a sale weekend or a late delivery can bring a wave of calls when your team has no spare time. Your call log shows whether that holds at your company.
A retail line carries lookups, calls that follow a rule you have written down, and calls that need a person. Sorting the line shows the split before you buy anything.
Three questions to sort the line by
Decide what goes into the count first. Three kinds of question go into it, and each asks something different of your systems.
The caller asks for a status and a date. Where a system holds both, a person reads them out. The question is a lookup.
Size, colour, variant, store, restock date. Stock data holds the answer and can change by the hour, so an out-of-date answer is worse than no answer.
A return is a process: the reason, the item, the label, the next step. It can go wrong when the caller hears something that contradicts your published policy.
Sort one sample of calls into four piles
Pull recordings or transcripts from several typical weeks, and from one promotion week if you have one. If your phone system does not record the reason for a call, listen to a sample and tag it by hand.
The caller asks where the order is or when it arrives. Note whether your order system held a date at the moment of the call.
The caller asks whether an item is available, in which variant and in which store. Note whether the stock data was current.
The caller wants to send something back. Note whether your returns policy covered the case.
A complaint, a damaged or missing parcel, a high-value order, a refund outside policy, an exception nobody has seen before.
Add a column and fill it in for every call: could the answer be read from a system at that moment? Count the calls marked yes.
Keep the sample. A vendor’s average describes other companies’ customers, and your own sample is the baseline for reading a pilot.
Read the sample before you pick a pile
- If most calls are order-status calls and the system holds a date for them, start with those calls.
- If many status calls end with no date in the system, the gap sits in the data and an agent cannot fill it.
- If stock calls dominate, check how often the stock data refreshes, because the agent repeats what the data says.
- If returns dominate, read your returns policy against the calls: a gap in it can produce escalations whoever answers.
- If most calls need a person, leave the line with people and look at staffing and routing first.
Treat the sample as a hypothesis and let a pilot test it. Every call comes back as a transcript, so the pilot shows how many callers asked for something the system did not hold. The AI receptionist page describes the inbound line the agent answers.
What the agent does on these calls
The page for retail and e-commerce sets out what Benerra would build for a retailer. Four behaviours matter for the sample above.
- It identifies the order, reads the current status and states the date your system holds, while the customer waits.
- Where the status is bad news, it says so, records the case and routes it.
- It checks stock by variant and by location at the moment of the call. Where an item is out, it offers the nearest store, the restock date if you hold one, or the closest variant.
- It opens a return inside the rules you configured and escalates anything outside policy with the reason written down.
Benerra confirms in writing which systems are connected before the agent takes a live call. Where a system cannot be read and written programmatically, Benerra says so during scoping. If your order feed is late or wrong, the agent repeats the error, so ask how often each feed refreshes.
Three things it will not do
The agent collects no card numbers, security codes or bank details by voice. A payment goes to a person or to a payment link.
The agent says there is no date yet and what happens next. A guessed date that proves wrong can bring the caller back with a complaint.
Refunds, discounts and exceptions to your returns rules carry money. The agent applies the rules you gave it and escalates the rest with the reason written down.
The article on what a voice agent should never be asked to handle lists four categories that belong to a person by design. The exit path covers how a call moves to a person.
Pilot one question and read the results
Benerra quotes the agent’s go-live timing per project and promises no number of days. Ask each vendor, including Benerra, how a limited slice of calls can be routed to the agent. Start with one question on that slice, then widen it.
Review five items from the pilot.
- The transcripts of the pilot calls, starting with the shortest and the longest.
- What the agent told the caller, whether a status and date, a stock answer or a return, against your system at the time of the call.
- The calls handed to people, with the reason for each: a problem status, a request outside policy, or a caller who asked for a person.
- The callers who rang again about the same order, a sign of whether the first answer held.
- If the pilot covers returns, the returns the agent opened, against the rules you configured.
Reading every call by hand takes more reviewers as volume grows beyond the pilot. Call analytics on the agent’s own calls is included in its package; ask which of these five checks it runs on your line. Locator, Benerra’s speech analytics, covers your human team’s calls: it scores 100% of recorded calls against a checklist you set. Manual QA in our own operations reviews under 5% of recorded calls.
The same question on the delivery side
A delivery operation gets the same question about a parcel in transit. The page on delivery and logistics describes the calls on that side: confirming a window, recovering a failed delivery and answering where a parcel is. Confirmation and recovery calls are outbound calls: before the first one, ask your counsel which rules apply in each market you call.
Steps for this week
- Pull the call records for several typical weeks and one promotion week if you have one.
- Tag a sample of calls into the four piles and mark each call the system could have answered.
- Check how often the order and stock data refresh, and who owns each feed.
- Write the escalation rules in your own words: what goes to a person, and what the caller hears.
- Pick one of the first three piles for the pilot and keep the sample as the baseline.