Customer logo
Case study

100 notification emails a day? Emma handles them now.

As the first TradeLink customer, WEKO Wohnen is handing over its delivery notification inbox to AI Agent Emma: over 50% of emails processed without intervention, taking about a minute per task instead of 10–15 hours of inbox work per week.

Header visual
From live operations
10–15 hours/week

time spent on inbox management before Emma

~1 minute

from email to processed transaction

> 50 %

of notification emails processed without manual intervention

WEKO Wohnen is one of the top 15 companies in the German furniture retail sector: family-run, founded in 1948, with around 1,200 employees and 5 logistics locations. 80 to 120 manufacturers and freight forwarders deliver regularly, and up to 50 delivery teams are on the road every day.

Since June 2024, WEKO has managed deliveries at its Rosenheim and Pfarrkirchen locations using TradeLink's time slot management system, handling around 800 bookings per month. However, a persistent challenge remained: many manufacturers do not book slots on the platform themselves, but instead send emails. Some send one per delivery, some six to ten for a single truck, and some provide no time at all.

The situation before Emma

The goods receipt team spent two to three hours a day on the notification inbox alone—10 to 15 hours per week. Processing delivery notes took just as much time: working through folders, searching for or creating appointments in the calendar, manually attaching delivery notes, printing everything out, and preparing physical folders for the next day. Canceled deliveries were only noticed when a prepared appointment suddenly went missing.

"This inbox basically contains everything. It's not just 'I'm notifying you and arriving on this day,' it's all sorts of things. That fills up the day."
Francesco Autorino, Head of Logistics, WEKO Wohnen

The goals

  • Automate the daily flood of delivery notifications in goods receipt, which can reach up to 100 emails per day.
  • A calendar as a reliable source of truth instead of Outlook folders, printouts, and paper files.
  • Relieve the team without losing control: routine tasks go to the agent, while decisions remain with the people.

Going where the information already is

Instead of forcing all partners onto the platform, Emma retrieves information where it arrives anyway: in the email inbox. In February 2026, WEKO became a pilot customer. For Autorino, this was about more than just efficiency:

"If we can pull this off—emails come in, the AI processes them—that would be a milestone. Not just for us: most companies in the industry are still working with emails."
Francesco Autorino

Since the go-live, every email from WEKO's notification folder in Rosenheim is automatically copied to Emma and processed within about a minute. Specifically, the agent now performs the following:

  • Assigning delivery notes: Emma identifies reference and commission numbers, finds the corresponding delivery in the calendar, and attaches the documents—even if 20 delivery notes are contained in a single PDF.
  • Preparing notifications: If a manufacturer reaches out only via email or fax, Emma creates a calendar proposal based on the specific day, time frame, or slot. The appointment is only created once the team confirms it.
  • Processing changes: Emma reliably detects delays and adjusts the appointment accordingly (e.g., "one hour later"). When it comes to changes in the cargo itself, such as partial deliveries ("8 instead of 9 upholstered furniture items"), detection is not yet at the target level; we are working on this.
  • Detecting duplicate notifications: If both a manufacturer and a carrier report the same goods, Emma uses the references to recognize that it is the same delivery and updates the record with the most recent information.
  • Identifying irrelevant content: Emma marks pure tracking and shipping emails as irrelevant, so the team no longer needs to read them individually. The team currently moves these to the archive themselves, but automatic filing is already planned.

Humans remain in the driver's seat

A new delivery is never created without human confirmation: Emma prepares the data, and the team provides manual approval. In the AI Task Manager, the team can see at a glance what Emma has completed (green), what is currently in progress (yellow), and where she needs help (a red dot, essentially like an unread message). The team provides feedback with a thumbs up or down, which the agent uses to learn.

The approach: prove it first, then take over

In May 2026, the test began with 32 real WEKO emails: Emma solved two-thirds correctly right away, and after the first improvements, that figure rose to over 80 percent. In mid-June, WEKO-IT activated automatic forwarding; when the team's manual preparatory work was removed, the success rate initially dropped. That is exactly what the test run was for: within a few days, Emma was again processing more than half of all live emails independently, without needing to consult the team.

At the beginning of July, the decision was made together with the team that works with it every day:

"Yeah, let's do it. There's no other way than just trying it out."
Francesco Autorino

Since July 6, 2026, at 8:00 a.m., Emma has been live—the first productive deployment of a TradeLink AI agent. What the team notices flows directly back into the product: the requested notification for cancellations was live one week later.

"So far, I'd say: good. We've even praised Emma already."
Antigona Cengiz, Goods Receipt Rosenheim, WEKO Wohnen, after the first week of live operation

What changes for the team

  • An inbox that sorts itself: Instead of working through every email one by one, the team works in the task manager. The goal: only the cases marked in red require attention.
  • Nothing gets lost anymore: Cancellations are actively flagged instead of silently disappearing from the calendar. Follow-up questions from tour planning ("When did the manufacturer provide notice?") are answered immediately, without searching through deleted emails.
  • No more busywork: Emma automatically splits bulk PDFs into individual documents.
  • The team helps shape the process: Feedback via thumbs up or down goes directly into development; some user requests have been implemented within a week.
  • And no one loses their job: The time saved is redirected toward tasks that were previously left undone.

The result

  • Faster: The process is completed about one minute after the email arrives. Follow-up questions are answered immediately because emails and delivery notes are attached directly to the delivery.
  • More cost-efficient: The process previously tied up 10 to 15 hours of inbox work per week. The time saved is redirected toward neglected tasks: higher output with the same team, growing without needing new hires.
  • Better: Standardized reference numbers make the calendar a reliable planning tool. No silent cancellations, no lost documents, a more satisfied team.

"That’s why I’m not cutting this position now — I have so many open tasks that I can shift to other things."
Francesco Autorino, Head of Logistics, WEKO Wohnen

What happens next

Emma is live, but she’s not finished yet. The next steps are clearly defined: reaching the target accuracy of 90–95%, reliably detecting changes in cargo, automatically filing irrelevant emails, bundling shipments from manufacturers that send six to ten individual notification emails, and switching from active verification to exception-based notifications. After that, the remaining WEKO logistics sites will follow. And within the purchasing association of around 200 stores, most are still working with email inboxes today—the very process that WEKO is currently leaving behind.

And Employee of the Month? Not yet, but the team was already praising Emma during her first week on the job. For a colleague who had her first day just seven weeks ago, that’s not a bad start.

"That doesn't mean I'm cutting staff—I have so many other open tasks that I can shift those resources to instead."

Francesco Autorino
Head of Logistics

Discuss your use case with us

We will show you how TradeLink and our AI agents can handle your specific processes.