Agents offered an add-on unprompted in almost 8 of 10 order calls

The share of order calls in which the agent offered an add-on unprompted rose from 60.5% to 77.8% in two months. With Lansy AI, the team set up scoring for each type of customer call and focused on two actions in the conversation: offering an add-on and reading the order back.

+29%

order calls with an unprompted add-on offer

4.3×

more often the agent reads the order back before the total

+15%

order calls where the agent offers cutlery and condiments

Pilot from June 16 to August 31, 2026. Change from the start of the pilot to its end.

Takeout order ready for pickup

A sushi delivery service

Industry
Sushi delivery
In business
Since 2011
Pickup points
3
In the pilot
Call center and pickup points

From the day the company opened, talking to the customer has been part of the service: confirm the order, offer a fitting add-on, and answer carefully when something goes wrong.

The agents already had rules for these conversations. In the pilot, the company wanted to make add-on offers more regular and see how the rules held up in everyday calls. Lansy AI reviewed more than six thousand call center conversations.

Listening to a sample of calls rarely caught an order

Calls used to be reviewed rarely and by sample. A manager could spend the time on a call about a courier or a delivery change and never hear how an agent takes an order and offers add-ons.

Reviewing every call showed why. Order calls were less than a third of the flow. The rest were about delivery status, questions, complaints, and changes to orders already placed.

This mattered for scoring the agents too. A customer could call after ordering in the app, with the sauces already chosen. Listing the sauces again only because the standard says so would be wrong. A call from an unhappy customer needed its own criteria: understand the problem and help solve it.

Two actions in every order call

Lansy AI was connected to the call recordings the company already kept, and compact microphones went to the pickup points. The team had the first reviews within days. During the pilot the criteria were tuned to the company’s processes.

Each type of call got its own scoring rules. For order calls the team focused on two actions: read the order back before naming the total, and offer one fitting add-on.

That gave the agent a clear focus. The offer had to fit the order. Listing several items at once could turn into a sales patter that is easier for the customer to refuse.

New goals were tested separately. The operations director asked for a specific item to be offered more often, and Lansy AI showed in how many calls it was offered and how often customers agreed. While the team tested a goal like this, it did not affect the agent’s score.

The next day the agent sees the review of their own call

The review shows the criteria met, specific episodes, and recommendations. The agent can go back to the exact moment in the recording and work out what to do differently.

That gave material for small changes in the next call: read the order back, offer the add-on in time, ask the question more politely. The team could return to the same actions and see how they were carried out.

LSushi delivery service›Conversations

Roll order, switched to pickup

Call center · order call · recording 04:30

Overall score

20 /21

Greeting

2 / 2

Needs discovery

2 / 2

Order taking

2 / 2

Upsell

2 / 2

Order processing

5 / 5

Closing

1 / 2

Language and impression

6 / 6

Summary

The customer called to order rolls and switched from delivery to pickup during the call. The order was accepted, and the items, the total, the ready time, the pickup point, and the loyalty points were agreed. The agent’s strength: confirming the order step by step, offering a drink, and processing the switch to pickup correctly, with the discount and the points. To work on: before the final goodbye, check whether the customer has any questions left.

Greeting · 2 / 2

Greeted the customerdone
Introduced themselves (name and brand)done

Key moments

  1. 00:01 – 00:17Greeting: the agent named the company and introduced themselves“[The company], my name is [agent’s name], hello.”
  2. 04:10 – 04:28Closing: the agent did not check whether the customer had questions left“All good. We’ll see you at the pickup point at 7:45 p.m. Goodbye.”

Before closing, the agent did not check whether the customer had any other questions.

Better approach

Before the final goodbye, add: “Do you have any other questions about your order?”

A review of one order call: 20 of 21 criteria. The agent read the order back, offered a drink, and switched delivery to pickup. Before the goodbye they did not ask whether the customer had questions left, and that became the topic of the next review.

An agent could disagree with a score and leave an objection right in the report. The Lansy AI team reviewed these objections and refined the criteria. That way the criteria quickly caught up with situations the general rules described too loosely.

By August the add-on offer was heard more often, and pickup errors halved

In June agents offered an add-on unprompted in six order calls out of ten. In August, in almost eight out of ten. The offer is now heard a third more often.

Agents began reading the order back four times as often: from 14% of conversations in the first week of observation to 60% in the last full week of August. Both figures describe the two actions the team chose at the start of the pilot.

The two actions in order calls

Share of order calls

Agent offers an add-on unprompted

60.5%77.8%

Agent reads the order back before the total

14%60%

Add-on offer: June against August. Order read back: first week of observation against the last full week of August.

Every day the team works from the same pages

Besides the call reviews, each team member has a page with the week’s numbers and a training plan for the next week. The pickup points have a shared dashboard. Both stayed in use after the pilot; the pages below are from the first week of September.

LSushi delivery service›Team

Pickup point team member

Employee page · August 31 – September 6, 2026

Kept a calm, friendly tone in 99% of conversations, above the average for the location and the chain. Upsell rose to 5.5/10, and communication quality stays the main strength at 9.9/10. A line that works: “Wasabi, ginger, soy sauce, need anything?” The customer hears clear options and keeps the choice.

Visits

113

−39% vs last week

Average score

7/10

order taking is slipping

Upsell conversion

65%

+11.4 pp vs last week

Sauces offered

80%

+2.2 pp vs last week

Skills

Consistently strong

Calm, friendly tone

Growing

A drink, mochi, or snack with every new orderRead the order back before paymentNeutral reply to off-topic questions
Employee page: the week’s numbers for one team member and the training plan for the week, what the person sees about their own work.
LSushi delivery service›Dashboard

Visits at pickup points

1,698

for the period · September 1–8

Visits by type

  • 83849%Ready order pickup
  • 56433%Order at the counter
  • 18211%Complaint or request
  • 1006%Extra item purchase
  • 121%Handoff to a courier
  • 20%Phone order

Score distribution

Problematic · 19%325
Normal · 71%1,213
Excellent · 9%160
Pickup point dashboard: the overall picture across locations, how many visits and for what reasons, and how the scores were distributed.
“I can see the numbers improving. The team keeps an eye on them. The tone of the conversations has improved a lot, no question about it.”
Founder of the company

Goals by location and a link to sales are next

01

Goals by team and by location

Pickup points sell differently. Goals will account for the differences in orders and in how customers behave.

02

Conversations linked to sales

Connect the changes in calls to actual sales, to see which offers bring results.

See how Lansy AI reviews your team’s conversations

In the demo, Lansy AI reviews conversations against your criteria and picks out the moments worth discussing with the team.